Scribd
Hochladen
10 Ansichten44 Seiten

Cs

Das Dokument ist ein Lehrplan für Computational Intelligence und behandelt verschiedene Themen wie Fuzzy Logik, Evolutionäre Berechnung, Genetische Algorithmen, natürliche Sprachverarbeitung und künstliche Immunsysteme. Es enthält Fallstudien zu intelligenten Abfallklassifizierungssystemen, Objekterkennungsrobotern und der Anwendung von Genetischen Algorithmen zur Lösung des Traveling Salesman Problems. Der Lehrplan hebt die Bedeutung von Computational Intelligence bei der Lösung komplexer realer Probleme hervor und beschreibt verschiedene Ansätze und Techniken.

Hochgeladen von

O50Tejas Pote
Copyright
© All Rights Reserved
Wir nehmen die Rechte an Inhalten ernst. Wenn Sie vermuten, dass dies Ihr Inhalt ist, beanspruchen Sie ihn hier.
Verfügbare Formate
Als PDF herunterladen oder online auf Scribd lesen
10 Ansichten44 Seiten

Cs

Das Dokument ist ein Lehrplan für Computational Intelligence und behandelt verschiedene Themen wie Fuzzy Logik, Evolutionäre Berechnung, Genetische Algorithmen, natürliche Sprachverarbeitung und künstliche Immunsysteme. Es enthält Fallstudien zu intelligenten Abfallklassifizierungssystemen, Objekterkennungsrobotern und der Anwendung von Genetischen Algorithmen zur Lösung des Traveling Salesman Problems. Der Lehrplan hebt die Bedeutung von Computational Intelligence bei der Lösung komplexer realer Probleme hervor und beschreibt verschiedene Ansätze und Techniken.

Hochgeladen von

O50Tejas Pote
Copyright
© All Rights Reserved
Wir nehmen die Rechte an Inhalten ernst. Wenn Sie vermuten, dass dies Ihr Inhalt ist, beanspruchen Sie ihn hier.
Verfügbare Formate
Als PDF herunterladen oder online auf Scribd lesen
SYLLABUS Unit I: Introduction To Computational Intelligence rie am Exemplar/Case Studies: Study of Inteligent Waste Classification System using Computational Intelligence. Unit: Fuzzy Logic SETA Introduction to Fuzzy Set: Introduction, definition, membership Function, Fuzzy operator, Fuzzy Set Characteristics, Fuzziness and Probability. Fuzzy Logic and Reasoning: Fuzzy Logic: Linguistics Variables and Hedges, Fuzzy Rules Fuzzy Inferencing: neuro inferencing Fuzzifcation, Defuzzfication Fuzzy logic Controllers: Fuzzy logic Controllers, Fuzzy logic Controller Types. Exemplar/Case Studies: Study of Object Detection Robot Using Fuzzy Logic Controller. Unit II: Evolutionary Computing (06 hrs) Introduction, Evolutionaty Computing, Terminologies of Evolutionary Computing, Genetic Operators, Evolutionary Algorithms: Genetic Algorithm, Evolution Strategies, Evolutionary Programming, Genetic Programming, Performance Measures of EA, Evolutionary Computation versus Classical Optimization, ‘Advanced Topics: Constraint Handling, Multi-objective Optimization, Dynamic Environments. ‘Swarm Intelligence: Ant Colony Optimization, Exemplar/Case Studies: Study of Engineering application of Artificial humming bird algorithm, Unit IV: Genetic Algorithm Zi a (06 hrs) Introduction to Basic Terminologies in Genetic Algorithm: Individuals, Population, Search space, Genes, Fitness function, Chromosome, Trait, Allele, Genotype and Phenotype. GA Requirements and Representation: Binary Representations, Floating-Point Representations Operators in Genetic Algorithm: Initialization, Selection, Crossover (Recombination), Mutation; fitness score, Stopping Condition, reproduction for GA Flow, Constraints in Genetic Algorithms. Genetic Algorithm Variants: Canonical Genetic Algorithm (Holland Classifier System), Messy Genetic Algorithms, Applications, and benefits of Genetic Algorithms, Exemplar/Case Studies: Use Genetic Algorithm to design a solution to the Traveling Salesman Problem. Solution: 1. Use Permutation Encoding 2. Define Objective Function. 3. Apply Selection Method 4. Crossover 5. Mutation 6. Repeat Until stopping criteria is met. [Link]. ote iereceaH seeta Unit V: Computational IntelligenceandNLP (06h), Introduction, Word embedding Techniques-8ag of Words, TF-IDF, Word2Vec, Glove, Neural word embedding, Neural Machine Translation, Seq2Seq and Neural Machine Translation, translation Metrics (BLEU Score & BERT Score), Traditional Versus Neural Metrics for Machine Translation Evaluation, Neural Style Transfer, Pertained NLP BERT Model and its application. Exemplar/Case Studies: (1) Study of Patient Triage using ChatGPT wi diagnoses. (2) Study of Question Answering System with BERT. can be utilized by physicians for expedited s (06 hrs) ‘Unit VI: Artificial Immune Systems i Natural Immune System, Artifical Immune Models, Artificial Immune System Algorithm, Classical View Models, Clonal Selection Theory Model, Network Theory Model, Danger Theory Model, Dendritic cell Model, Applications of AIS models. Exemplar/Case Studies: Study of an artificial immune system with bootstrap sampling for the diagnosis of recurrent endometrial cancers. VIKAS 26 27 28 253. Differences Between Fuzziness and Probability Fuzzy Logic and Reasoning 261 Fuzzy Logic Fuzzy Inferencing System (FS) 27.1 Functional Blocks of FIs 27.2 Methods of FIS 273 Neuro Inferencing 274 Fumzification and Defuzzification Fuzzy logic Controllers (FLC) 28.1 Introduction of Fuzzy Logie Controllers 28.2 Architecture and Operations of FLC System 283 Applications of FLC System 284 Fuzzy Logic Controller Types Case Study : Study of Object Detection Robot Using Fuzzy Logic Controller Exercise Unit II: Evolutionary Computing Ba Introduction 351.1 Evolutionary Computing 312 Terminologies of Evolutionary Computing 313 Genetic Operators 32 Evolutionary Algorithms 321 Genetic Algorithm 322 Evolution Strategies 3.23 Evolutionary Programming 33. Genetic Programming 331 Introduction 33.2 Working of Programming 333 Representation of Programming 334 Types of Genetic Programming 335 Applications of Genetic Programming 34 Performance Measures of EA 341 Evolutionary Computation Versus Classical Optimization 35 Advanced Topics 351 Constraint Handling 352 Multi-objective Optimization 3:53 Dynamic Environments 3.6 Swarm Intelligence 364 Ant Colony Optimization + Case Study : Study of Engineering Application of Artifical Humming Bird Algorithm + Exercise Unit V: Genetic Algorithm” 27 27 27 230 210 2aa 232 212 213 233 214 245 216 216 218 3.13.20 BL Ba 32 33 33 33 38 38 39 39 Bal 3a 312 32 312 313 314 314 35 3.16 317 317 318 CONTENTS em (Seve ness Unit E introduction to Computational Intelligence 129 11 Introduction to Computational intelligence qa 1.2 Paradigms of Computational intelligence hi 121 Fuzzy Logic an 1.22 Neural Networks 13 123 Evolutionary Computation i 1.24 — Learning Theory 7 robabilstic Methods 13 biclel one aie Intelligence and Computational Intelligence 16 13:1 Whats Artificial Intelligence? 16 ‘at is Computational Intelligence? 1s i cote of pete Intelligence (Al) and Computational intelligence (C1) v 14 Approaches to Computational Intelligence wW 141 Genetic Algorithms uv 142. Swarm Intelligence . an 143 Ant Colony Optimization (ACO) 13 144 — Rough Set Theory 1s 14S Expert Systems 116 1.4.6 Hybrid Systems 116 1.47 Machine Learning 116 1S Synergies of Computational Intelligence Techniques 116 16 Applications of Computational Intelligence 118 17 _ Grand Challenges of Computational Inteligence 1s + Case Study : Study of Inteligent Waste Classification System using Computational Intelligence 119 + Exercise 120 Unit I: Fuzzy Logie Zi TE aaa 21 Introduction to Fuzzy Set 21 211 Fuzzy Numbers 21 212 _ Definition of Fuzzy Set 22 22 Membership Function 24 221 Introduction of Membership Function 24 222 _ Definition of Fuzzy Set Membership Function 24 223 Types of Membership Functions 24 2.24 — Applications of Fuzzy Set Membership Functions 24 23° Fuzzy Operator 25 2.3.1 — Operation on Classical Set 25 23.2 — Properties of Classical Set 25 24 Fuzzy Set Characteristics 26 2.5 Fuzziness and Probability 26 25. Introduction of Fuzziness and Probability oe 252 Relationship Between Fuaziness and Probability 57 42 Introduction of Genetic Algorithm (G-A) 4 411 — Basic Structure of GA 4 412 Psedo-Code ‘ 42 Introduction to Basic Terminologies in Genetic Algorithm aI 42:1 Individuals s 422 Population a 423. Search Space a 424 Genes 4 425. Fitness Function bt 426 Chromosome i 427 Trait 7 428 Allele 7 429 Genotype and Phenotype re 43. GARequirements & Representation is 43.1 Binary Representations a 432. Floating-Point Representations 7 44 Operators in Genetic Algorithm a 444 Initialization a 442 Fitness Assignment Operator 7 443. Selection aa 444 Crossover (Recombination) ia 445 Mutation iy 446 Fitness Score ay 447 Stopping Condition oat 448 — Reproduction for GA Flow ae 449 — Constraints in Genetic Algorithms ia 45 — Genetic Algorithm Variants ae 45.1 Introduction aus 452 Canonical Genetic Algorithm 414 453 Holland Classifier System ian 41s 45.4 Messy Genetic Algorithms (mGAs) 45.5 Applications and Advantages of Genetic Algorithms 416 * Case Study : Use Genetic Algorithm to design a solution to the Traveling Salesman 418 419 + Exercise ‘Unit V: Computational Intelligence and NLP 2 5.1 Introduction 5.1.1 Foundations of Computational Intelligence a 51.2 Key Applications of Computational Intelligence in NLP = 5.13. Challenges and Future Directions = 52 Word Embedding Techniques ae 5.2.1 Bag of Words ee 522 TEIDF 3 58 5.3 Word2Vec SwaAw 54 55 56 87 58 61 62 63 init VE Artificial Immune Systems 53.1 Why to use Word2Vec? 532 Working of Word2Vec 533 Word2Vec Architecture 534 Relation Between Word2Vec and NLTK 53,5 Activators and Word2Vec 536 What is Gensim? Glove 5.41 Implementation of GloVe in Python - Neural Word Embedding 55.1 Creating a Custom Set of Word Embeddings 5.5.2. Using Pre-Built Set of Word Embeddings 55.3 Creating Word Embeddings on the Fly Neural Machine Translation 5.61 Explanation of NMT and Encoder Decoder Structure 5.62. Cross-Entropy Loss ‘Seq2Seq and Neural Machine Translation 57.1 Data Collection and Cleaning 5.7.2. Input Creation for Encoder and Decoder 57.3 Model Building 5.74 Training Model 5.75 _ Inference Model Building Translation Metrics 5.8.1 Traditional Versus Neural Metrics for Machine Translation Evaluation 5.82 Neural Style Transfer 5.83 Pertained NLP BERT Model 5.84 Applications of BERT Model Case Study 1 : Study of Patient Triage using ChatGPT which can be utilized by physicians for expedited diagnoses Case Study 2 : Study of Question Answering System with BERT Exercise Natural Immune System 6.1.1 Key Components of the Natural Immune System 61.2 _ Functions and Mechanisms Antficial Immune Models 621 Key Components of Artifical Immune Models 622 _ Principles of Artificial Immune Models 623 Applications of Artificial Immune Models 624 Hybrid Models and Future Directions Antfcial Immune System Algorithm 63.1 Key Components of Artifical Immune System Algorithms 59 59 540 san saz 513 57 5a7 518 519 52. 522 523 5.24 528 5.23 5.29 531 532 534 535 536 537 337 537 5238 5.38 539 5.40 61-618 61 61 61 62 63 63 63 64 64 64 SVAL Working Principles of Artificial Immune System Algorithms 632 633 Applications of Artificial Immune System Algorithms 64 Classical View Models 641 Key Components of Classical View Models 642 Applications of Classical View Model 643 Waterfall Model 644 V-Model 645 Spiral Model Clonal Selection Theory Model (CST) 6s 65:1 Key Concepts of Clonal Selection Theory Model 65.2 Working Principles of Clonal Selection Theory Model 653 Applications of Clonal Selection Theory Model 654 — Future Directions and Innovations 6.6 Network Theory Model Social Network Theory 661 662 Electrical Network Theory 663 Network Theory in a Broader Context 6.7 Danger Theory Model 67.1 _ Key Concepts of Danger Theory 672 Working Principles of Danger Theory 673 Applications of Danger Theory 67.4 — Future Directions of Danger Theory 68 Dendritic Cell Mode! Key Concepts of the Dendritic Cell Model 681 68.2 — Working Principles of the Dendritic Cell Mode! 683 Applications of the Dendritic Cell Model 684 Future Directions of the Dendritic Cell Model 69 Applications of AIS Models Case Study : Study of an Artificial Immune System with Bootstrap Samy Endometrial Cancers . Exercise * Model Question Papers > In-Sem. Exam. (30 Marks) > End-Sem. Exam. (70 Marks) REK [unr -y) INTRODUCTION TO COMPUTATIONAL INTELLIGENCE a , 1.1 INTRODUCTION TO COMPUTATIONAL INTELLIGENCE Computational Intelligence is sometimes also referred to as soft computing, which is a specific field of study where the task is to make computers learn some real-life or complex problems from the experimental data or observations. In computational intelligence, there is a set of approaches ‘or methodologies used to address real-life or complex problems. Generally, it is impossible to solve real-life problems using traditional computing methods because of complexity, uncertainty or problems that do not have a proper definition. So, it is complex and not feasible to convert these problems into computer understandable binary format (© and 1). In such cases, we have computational intelligence with which it is possible to formulate these problems into computer understandable format, Computational Intelligence uses some techniques that try to imitate the human's way of questioning and reasoning, What is Computational Intelligence? Computational intelligence is a way to imitate the reasoning process of human intelligence and accomplish this CI approaches which are used in combination. To achieve human intelligence is complex, but the way it reasons or questions the problem at hand can be replicated and the same is done using Cl approaches. The approaches which are used in Cl are as follows: Neural Networks Learning Theory + Fuzzy Logic . ‘+ Evolutionary Computation * ‘+ Probabilistic Methods Use of Computational Intelligence + As we have seen, computation intelligence has the ability to consider real-life complexity and then make probabilistic decisions, which makes it very useful in cases of scheduling industry procedures, Disease diagnostics, Video games visualization, Translation systems (ike Alexa, Siri which can understand human command and follow along precisely), intelligent robots, infobots autonomous vehicles etc. WKAS + Computational intelligence is a growing field and has some interesting approaches like evolutionary computation and learning theory, making it competent enough for a wide variety of real-life problems ‘Advantages of Using Computational Intelligence We, as human being, are working on evolution from ancient time and one of the goals we have is creating an intelligent system which can make our life easier. This goa! is progressive with time, our requirement changes and we need something better. Example of Cell Phone: ‘+The primary purpose was to convey information from cone place to another so it started with sending ravens (birds with messages), then we got telegraphs and then fax. The revolutionary discovery was telephones, With time our requirements grew because we wanted more comfort. + So, the size was optimized the way of installation was optimized and then we have portable phones, which are nothing but our mobiles now. + These portable phones were only able to make calls, then we wanted to have a text feature and a way to add all the contacts and then utilities like calendar, watch, alarm, games etc. but we wanted something more our goal shifted further. ‘+ Now we have smartphones, which are still getting new features a day in and day out. To Cop-up with such kind of goal, we need to have an approach that can keep up with evolution where Computational Intelligence advantages over _ other approaches. 1.2. PARADIGMS OF COMPUTATIONAL INTELLIGENCE 1.2.1 Fuzzy Logi + As the name suggests fuzzy or unclear, there are @ number of instances in our day to day life where the situation is not clear, whether to say yes or no, may be its yes/no or no/yes (e.g. asking girlfriend if she wants to buy an expensive dress she may say no with a weird face, man is supposed to read between the lines it is yes). Fuzzy logic decides final output based on the levels of possibilities. aa COMPUTATIONAL INTELLIGENCE For example, Yes sure/ Yes may be/ Not sure / may be no / No. Basically, fuzzy logic is the conditions of ifs and else, which is easy to understand but is not very accurate. It is used for control systems, intelligent appliances, voice recognition etc. The term fuzzy refers to things that are not clear or are vague. In the real world many times we encounter 2 situation when we can not determine whether the state is true oF false, their fuzzy logic provides very valuable flexibility for reasoning. In this way, we can consider the inaccuracies and uncertainties of any situation. Fuzzy logic is a form of many-valued logic in which the truth values of variables may be any real number between 0 and 1, instead of just the traditional values of true or false. It is used to deal with imprecise or uncertain information and is @ mathematical method for representing vagueness and uncertainty in decision-making Fuzzy logic is based on the idea that in many cases, the concept of true or false is too restrictive and that there are many shades of gray in between. It allows for partial truths, where @ statement can be partially true or false, rather than fully true or false. Fuzzy logic is used in a wide range of applications, such as control systems, image processing, natural language processing, medical diagnosis and artificial intelligence, The fundamental concept of Fuzzy logic is the membership function, which defines the degree of membership of an input value to a certain set or category. The membership function is a mapping from an input value to a membership degree between 0 and 1, where represents non-membership and 1 represents full membership. Fuzzy logic is implemented using Fuzzy Rules, which are if-then statements that express the relationship between input variables and output variables in a fuzzy way. The output of a Fuzzy Logic system is a fuzzy set, which is a set of membership degrees for each possible output value, In summary, Fuzzy logic is a mathematical method for representing vagueness and uncertainty in decision ‘making, it allows for partial truths and it is used in a wide range of applications. It is based on the concept of membership function and the implementation is done using Fuzzy Rules. a2 INTRODUCTION TO COMPUTATIONAL wey t + In the boolean system truth value, 10 rep false value but in the fuzzy system, there ig nok Sole the absolute truth and absolute fase vajye yo In fy logic, there isan intermediate value too presen is partially true and partially false. ‘hich [> Yos/1 Is itcold? | Boolean lye L+ novo (> Very much /0.9 Isit cold? Fille /0.25 —bFuzayioje Ls Very less /0.1 Fig. Architecture Its architecture contains four parts: 1. Rule Base: It contains the set of rules and the I-THE conditions provided by the experts to govern the decision-making system, on the basis of linguistic information, Recent developments in fuzzy theory offer several effective methods for the design and tuning of fuzzy controllers. Most of these developments reduce the number of fuzzy rules. 2. Fuzzification: It is used to convert inputs ie. cisp numbers into fuzzy sets. Crisp inputs are basicaly te exact inputs measured by sensors and passed into te control system for processing, such as temperatus, pressure, 1pm's, ete. 3. Inference Engine: It determines the matching deste® of the current fuzzy input with respect to each rule ard decides which rules are to be fired according tot input field. Next, the fired rules are combined to for the control actions. 4. Defuzzification: It is used to convert the fuzzy #6 obtained by the inference engine into 2 cisp vee There ae several defuzztication methods aalebe 3" the best-suited one is used with a specific © system to reduce the error. Banal Input—| COMPUTATIONAL INTELUGENCE __ HH ‘Membership Function Definition: & graph that defines how each point in the mapped to membership value between 0 input space i ‘and 1. Input space is often referred to as the universe of Giscourse of universal set (u), which contains all the possible elements of concern in each particular application. There are largely theee types of fuzzifiers 1° Singleton fuzzifer 2. Gaussian fuzzifier 3. Trapezoidal or triangular fuzzifier What is Fuzzy Control? + This a technique to embody human-like thinkings into a control system. + Itmay not be designed to give accurate reasoning but itis designed to give acceptable reasoning, «It can emulate human deductive thinking, that is, the process people use to infer conclusions from what they know. + Any uncertainties can be easily dealt with the help of fuzzy logic. Advantages of Fuzzy Logic System + This system can work with any type of inputs whether itis imprecise, distorted or noisy input information. +The construction of Fuzzy Logic Systems is easy and understandable. © Fuzzy logic comes with mathematical concepts of set theory and the reasoning of that is quite simple. + It provides @ very efficient solution to complex problems in all fields of life as it resembles human reasoning and decision-making + The algorithms can be described little memory is required. Disadvantages of Fuzzy Logic Systems + Many researchers proposed different ways to solve a given problem through fuzzy logic which leads to ambiguity. There is no systematic approach to solve a given problem through fuzzy logic. ‘+ Proof of its characteristics is difficult or impossible in most cases because every time we do not get a mathematical description of our approach. + As fuzzy logic works on precise as well as imprecise data so most of the time accuracy is compromised. Appl + Its used in the aerospace field for altitude control of spacecraft and satelites. little data, so ‘+ Ithas been used in the automotive system for speed Control, traffic control aay INTRODUCTION TO COMPUTATIONAL INTELUGENCE + It is used for decision-making support systems and personal evaluation in the large company business © it has application in the chemical industry for controlling the PH, drying, chemical distillation process «Fuzzy logic is used in Natural Language Processing and Various intensive applications in Artifical intelligence. + Fuzzy logic is extensively used in modem control systems such as expert systems © Fuzzy Logic is used with Neural Networks as it mimics how a person would make decisions, only much faster. It is done by Aggregation of data and changing it into more meaningful data by forming partial truths 2s Fuzzy sets. 1.2.2_Neural Networks The human intelligence system works on biological neurons; for Cl, we have an artificial neural network, ‘Artificial Neural Network (ANN) attempts to replicate human neural network, where each node of ANN is referred to biological neuron. ‘A biological neuron takes input through input cells called dendrites. In neuron, the information is processed with logic and reasoning (Synapses) the output will be given by cells called axons. The same is referred to as the input layer, activation function and output layer in an Artificial Neural Network (ANN), Input is given with some weight which further is evaluated ‘and adjusted with feedback and the activation function decides output based on input parameters, In Cl, it has wide applications in the field of classification problems, regression problems, association problems and pattern recognition problems. 1.2.3 Evolutionary Computation + Those who endure evolution have more chances of survival, a natural phenomenon that we have seen over history. We have also seen an evolution in our lives and how our way of living has evolved from our childhood to our present state. * In computational intelligence, some of these biological evolution theories, such as reproduction or mutation, are considered for making artificial intelligence more robust to deal with real-life problems. * Over a period of time, we have created a number of evolution theories and the same is applied to create evolutionary algorithms for eg, “The survival of the fittest” is the theory and fitness function is the part the genetic algorithm. 2 S aa ov) INTRODUCTION TO COMPUTATIONAL INTELGEyeg COMPUTATIONAL INTELLIGENCE + In computation intelligence, evolutionary computation is used for optimization problems (the goal is to optimise the present state) and progressive problems (the goal isto predict the future state) ‘+ Evolutionary computation is a sub-field of Artificial Intelligence (Al) and is used extensively in complex optimization problems and for _ continuous ‘optimization. Evolutionary computation is used to solve problems that have too many variables for traditional algorithms. ‘+ Computers performing evolutionary computing run such evolutionary algorithms as genetic algorithms, evolutionary programming, genetic programming and swarm intelligence models like ant colony optimization or patticle swarm optimization, ‘+ The computational models using evolutionary algorithms apply evolutionary processes in order to solve complex problems. These evolutionary processes are inspired by biological evolution theory. * Evolving algorithms use principles such as inheritance from previous successful generations and natural selection where the best solutions pass their traits on to the successive generations. Working of Evolutionary Computation * An initial batch of possible solutions is created with the start of an evolutionary computation. The solutions once tried are refined as weaker solutions are stochastically removed and small random changes are introduced to successive generations. | As the generations pass the solutions become increasingly refined. In the end the solutions produced by evolutionary computation can be tightly optimized, even though in the beginning, the approach is not understood. ‘+ Evolutionary computation is a branch of Artificial Intelligence and is used heavily for complex optimization problems and also for continuous optimization, ‘+ Evolutionary computation techniques are used to handle problems that have far more variables than what traditional algorithms can handle. ‘+ These computational models employ evolutionary algorithms that essentially use evolutionary processes for the purpose of solving such complex problems. They use evolutionary principles like inheritance from effective previous generation models and natural selection, where the traits from the most effective models and passed on to future generation models. Evolutionary computing Evolutionary | perateoiesal | programming Why do we use Evolutionary Computation? Since evolutionary computing has the ability to produce tightly optimized solutions for a wide range of problems, they are extensively used in computer science. There are even variants that are created and used specifically for particular data structures and families of problems. This branch of artifical intelligence is also employed in evolutionary biology for studying common aspects of general evolutionary processes. ‘Types of Evolutionary Algorithms £ Genetic Algorithms (GA) 65 Genetic Programming (GP) @ Bi Evolutionary Programming(EP) Evolutionary Svaleies(ES) Fig. 1.4 There are various types of evolutionary algorithms, Here are the most significant ones: 1. Genetic Algorithms (GA) 2. Genetic Programming (GP) 3. Evolutionary Programming (EP) 4. Evolutionary Strategies (£5) Many more evolutionary algorithms also exist. These include Gene Expression Programming, _Differenta Evolution, Learning Classifier Systems and Neuroevolution. 1. Genetic Algorithms (GA) + Genetic Algorithms ate the most popular type © evolutionary algorithms. They find solutions t problems as strings of numbers. Most of these sting: are binary, but the most effective ones tend to sho something about the problem in question. + These algorithms make use of operators like mutator and recombination. Sometimes they use operators together * One of the uses of genetic algorithms is selecting the right combination of variables to build @ pred’ model. Selecting the right subset of variables essentially a combinatory and optimization problem COMPUTATIONAL INTELLIGENCE ‘The advantage of genetic algorithms is that it makes it possible for the best solution to emerge from the best Of prior solutions. It improves the selection over time. The whole idea behind genetic algorithms is to combine the different solutions generation after generation so that it can extract the best genes or vatiables from each solution. It helps creating better fitted individuals Genetic algorithms are also used for hyper-tuning parameters, finding the maximum or minimum of a function or the search for a correct neural network architecture (Neuroevolution),Itis also used in feature selection. The idea of Genetic Algorithms (GA) is to generate a few random possible solutions that represent different variables and then combine the best possible solutions inan iterative process. The basic genetic algorithm operations are selection (picking the most fitted solutions in a generation), cross-over (creating two new individuals, based on the genes of solutions) and mutation (changing @ gene randomly in an individual. Genetic Programming (GP) Here, the solutions to problems are computer programs. The ability of these computer programs to solve computational problems is what determines their fitness. Genetic Programming (GP) is essentially an automatic programming technique which favors the evolution of computer programs that solve or at least approximately solve problems. It involves essentially ‘breeding’ programs by continuously improving an initially random set of programs. Improvements are made by stochastic variation of programs and selection in line with some predefined rteria for judging the quality of a solution. Programs of genetic programming systems essentially evolve to solve predescribed automatic programming and machine learning problems. In its essence, genetic programming is a heuristic search technique that is commonly called ‘hill climbing’. It involves searching for an optimal or at least a suitable program among the space of all programs. Evolutionary Programming (EP) This is not too different from Genetic Programming, However, in Evolutionary Programming, the programs that need to be optimized have a fixed structure, while the numeric parameters can evolve, as) Svau INTRODUCTION TO COMPUTATIONAL INTELUGENCE © This evolutionary algorithm paradigm was first used by Lawrence J. Fogel in 1960 in an attempt to use simulated evolution as a learning process seeking to create artificial intelligence. © He used finite-state machines as predictors and evolved them. Right now, evolutionary programming is a wide evolutionary computing dialect that has no fixed structure or representation. It is becoming increasingly difficult to differentiate evolutionary programming from evolutionary strategies. ‘©The main operator of evolutionary programming is mutation. In evolutionary programming, members of the population are seen as part of a specific species rather than members of the same species. Every parent generates an offspring by using @ (¥ + H) survivor selection 4, Evolutionary Strategies (ES) * Evolutionary strategies usually work by making use of self-adaptive mutation rates. They work with vectors of real numbers as representations of solutions. ‘© Evolutionary strategies are optimization techniques that are based on the ideas of evolution. They use natural problem-dependent representations and: mainly make use of mutation and selection as search operators. ‘+ The operators are applied in a loop, an iteration of which is known as a generation. The sequence of generations continues till a termination criterion is met. Most evolutionary algorithms work on a genotype level, but evolutionary strategies work on a behavioral level ‘+ Since the physical expression is coded directly, an individual's genes are not mapped to its physical expression. This approach is followed to give rise to 2 strong causality so that @ small change in the coding gives rise to a small change in the individual and a large change in the coding causes a large change in the individual, 1.2.4 Learning Theory Learning theory basically means the philosophy of learning, how a leamer absorbs information, process it and then retain it for further decision-making problems. In learning theory goal is to study/understand different learning techniques through which learning can occur. Cognitive learning is also part of learning theory. COMPUTATIONAL INTELLIGENCE (1.2.5 Probabilistic Methods The probabilistic methods are non-constructive and non deterministic methods that are used for providing the existence of an object. In simple words, if there is a collection of objects and are assigned with certain properties Now, if all the objects in this collection does not have one specific property and then one chooses a random object from this collection, for this object, the probability to have that property is zero for sure similarly, if we show that probability is less than one, which will prove that there exist one or more objects that do not possess that specific property 1.3 DIFFERENCE BETWEEN ARTIFICIAL INTELLIGENCE AND COMPUTATIONAL INTELLIGENCE + We have witnessed nothing sort of a revolution in the history of mankind as Artificial Intelligence (Al) becomes mainstream in our everyday life and augmenting human capabilities to solve some of the complex problems that troubled us for a really long time. ‘+ What was once a distant dream of the future is now the future of mankind! We are at a stage where human intelligence and machine intelligence coexist at the onset of a data revolution. Artificial Intelligence aims at emulating human intelligence on machines to make them think and behave like humans. «The human brain has evolved for thousands of years to its current state. However, traditional AI was not enough to serve the increasing demands of machine learning. The shortcomings of AI opened up new avenues for the non-conventional models which eventually have rise to a new discipline called Computational Intelligence (Cl) ‘+The development of CI techniques follows a different path than that of the Al. Let us take a brief look at the new computational model called the CI and how it is compared to the classical Al approach, 1.3.1_ Whats Artificial Intelligence? Artificial Intelligence, commonly abbreviated as Al, is the study of intelligent behavior and how to make machines do things at which humans are doing better. Alis one of the technological breakthroughs of this digital era which aims at emulating human intelligence on machines to make them think and behave like human beings. (.6)__ INTRODUCTION To ComPUTATIONAL a. ete 1 is based onthe ideo that human inteligenes replicated in computer programs. though creating integent machines ~ ores tha ae oe smarter than human beings ~ is not new bay etter a part of modem science with the rise of digital ye“ and proliferation of the internet Tote) * 19 the practical side, Al means to crea programs that perform tase awl as or 4th humans. In simple terms, AI is thurman int demonstrated by machines, te 1.3.2 What is Computational Intelligence? ‘Computational Intelligence, also referred to as with the capability of computational adaptation, tolerance and high computation speed. Its the adaptive mechanisms to enable oF feciitate inerge, behavior in complex and changing environments, Computational adaptation means the ability of a system ty adapt to the changes in its input and output instances Ty adaptive mechanisms include the following AI paradigns that exhibit an ability to learn or adapt 10 ney environments: Swarm Intelligence (SI, Artificial Neue Networks (ANN), Evolutionary Computation (EC), Artifics Immune Systems (AIS) and Fuzzy Systems (FS). All these «1 paradigms are the building blocks of computations intelligence. It is an emerging approach to computrg which parallels the remarkable and intelligent ability of te human mind. Definition of Artificial Intelligence vs Computation’ Intelligence '* Artificial Intelligence (Al) is the study of intelige: behavior demonstrated by machines as opposed toth natural intelligence in human beings. It isan ae2¢ computer science that is concerned with & development of a technology that enables a mach ‘or computer to think, behave or act in a more hums! way. ‘* Computational Intelligence (Cl), on the other hard more like a sub-branch of Al that emphasizes 0! design, application and development of linguist? motivated computational models. It is the S04 adaptive mechanisms to enable or facilitate teh? behavior in complex and changing environments Comparison of Goals + AL is one of the remarkable te" breakthroughs of the 21* century which & emulating human intelligence on machines them think and behave like human beings _—~ COMPUTATIONAL INTELLIGENCE The idea is to create intelligent machines which can exhibit intelligent behavior and which can think and learn like human beings. «The goal is to implement human intelligence in machines. The primary goal of Clis to understand the computational paradigms that make intelligent, behavior possible in natural or artificial systems in complex and changing environments. ‘Applications + Alis the simulation of human intelligence, as opposed to natural inteligence, by machines, particularly computer systems. The best real-life example of Al is probably machine learning, which refers to automated systems that are able to process large volumes of data in order to extract useful information © Other common applications of Al are speech recognition, handwriting recognition, optical character recognition, machine vision, natural language processing, big data solutions and so on. Al is used in almost every sector including finance, healthcare, social media, business, tourism, etc. + The real-life applications of CI include intelligent household appliances, medical diagnosis, banking and electronics, optimization applications, consumer industrial applications and so on. 1.3.3. Comparison of Artificial Intelligence (Al) jonal Intelli Table 1.1 ‘Astiticial Intelligence (Al) _| Computational Intelligence (Cl) ‘Al is the study of inteigent | Cl is the study of adaptive behavior demonstrated by | mechanisms fo enable or machines as opposed to the | facilitate inteligent behavior in natural inteligence in human | complex and changing beings. environments ‘The goa of lis to understand paradigms that make inteligent ‘behavior possible in natural or atfcal systems in complex and changing envionments and Computat ce (CI) “The idea is to create inteligent ‘machines which can exhibit infetigent behavior and which can think and learn ike human beings, ‘Common applications of Al are ‘speech recognition, handwriting recognition, optical character recognition, machine vision, natural language processing, big data solutions and soon, ‘The realife applications of Cl include inteligent household appliances, medical diagnosis, banking and consumer electronics, optimization applications, industrial applications and so on. an INTRODUCTION TO COMPUTATIONAL INTELUGENCE [1.4 APPROACHES TO COMPUTATIONAL INTELLIGENCE Computational Intelligence (Ch is a field of study that encompasses various techniques and methodologies inspired by natural inteligence to solve complex problems. There are several approaches within Computational Intelligence, each with its own set of techniques and applications. Here are some prominent approaches: 1.4.1 Genetic Algorithms (GA) ‘© Inspired by the process of natural selection, GAs involve evolving a population of candidate solutions ‘over multiple generations to find optimal or near- optimal solutions to a problem. * Genetic Algorithms (GAs) are adaptive heuristic search algorithms that belong to the larger part of evolutionary algorithms. Genetic algorithms are based on the ideas of natural selection and genetics. These are intelligent exploitation of random search provided with historical data to direct the search into the region of better performance in solution space. They are commonly used to generate high-quality solutions for ‘optimization problems and search problems. ‘© Genetic algorithms simulate the process of natural selection which means those species who can adapt to changes in their environment are able to survive and reproduce and go to next generation. In simple words, they simulate “survival of the fittest” among individual of consecutive generation for solving a problem. + Each generation consist of a population of individuals and each individual represents a point in search space and possible solution. Each individual is represented as a string of character/integer/float/bits. This string is analogous to the Chromosome. Foundation of Genetic Algorithms Genetic algorithms are based on an analogy with genetic structure and behaviour of chromosomes of the population. Following is the foundation of GAs based on this analogy ~ ‘+ Individual in population compete for resources and mate ‘* Those individuals who are successful fittest) then mate to create more offspring than others + Genes from “fittest” parent propagate throughout the generation, that is sometimes parents create offspring which is better than either parent. ‘© Thus each successive generation is more suited for their environment. 2 S COMPUTATIONAL INTELLIGENCE Search Space The population of individuals are maintained within search space. Each individual represents a solution in search space for given problem, Each individual is coded as a finite length vector (analogous to chromosome) of components These variable components are analogous to genes. Thus @ chromosome (individual) is composed of several genes (variable components). qiobooo (aooooo Chromosome Fitness Score ‘+ A fitness score is given to each individual which shows the ability of an individual to “compete”. The individual having optimal fitness score (or near optimal) are sought. ‘+ The GAs maintains the population of n individuals (chromosome/solutions) along with their fitness scores. The individuals having better fitness scores are given more chance to reproduce than others, The individuals with better fitness scores are selected who mate and produce better offspring by combining chromosomes of parents, O00000 Population Fig. 1.5 ‘+ The population size is static so the room has to be created for new arrivals. So, some individuals die and ao) INTRODUCTION TO COMPUTATIONAL NTELUGe, t get replaced by new arrivals eventually creat generation when all the mating opportunity of, population is exhausted. It is hoped that least fit die. " + Fach new generation has on average more “beng, genes than the individual (solution) of prt generations, Thus each new generations have bent’ “partial solutions” than previous generations, + Once the offspring produced having no signifean ditference from offspring produced by prevent Populations, the population is converged mrt algorithm is seid to be converged to a set of solution, for the problem, Operators of Genetic Algorithms Once the initial generation is created, the algorithm evolves the generation using following operators. 1, Selection Operator: The idea is to give preference to the individuals with good fitness scores and allen them to pass their genes to successive generations 2. Crossover Operator: This represents mating between individuals. Two individuals are selected using selection operator and crossover sites are choser randomly. Then the genes at these crossover sites art exchanged thus creating @ completely new individu (offspring), For example ~ old Porent + [APS ICIP EIS Offspring Paro 2 fF HA 5 Efe Fig. 1.6 3. Mutation Operator: The key idea is to insert random genes in offspring to maintain the diversity in the Population to avoid premature convergence. For example ~ Before mutation [F][G][H][B][C][D][E][A After mutation [F][G][m][B][e][o][EIN Fig. 1.7 The whole algorithm can be summarized as : L 2. Randomly initialize populations p Determir 3. Until convergence repeat: (@) Select parents from population (b) Crossover and generate new Population (©) Perform mutation on new population (d) Calculate fitness for new population Example Problem and Solution using Genetit Algorithms: Given a target string, the goal is to produce target string starting from a random string of the same length. In the following implementation, following analogies are made Characters A-Z, a-z, 0-9 and other special symbols ar considered as genes. COMPUTATIONAL INTELLIGENCE as) ‘A string generated by these characters i considered as “dwomosome/solution/Individual neters which differ from 5 score is the number of chi Fit seemters in target string at a particular index, So fahidual having tower fitness value is given more preference cr Python3 // Cee program to create target string, starting from 7/ condom string using Genetic Algorithm #include «bits/stder+.h» using namespace std: // Number of individuals in each generation #Hdefine POPULATION_SIZE 100 // Nolid Genes const string GENE: “abcdefghijkimnopqrstuvwxyzABCDEFGHIJKLMNOP"\ "QRSTUVWXYZ 1234567890, .~i:_IN"#%E/(7O$(E)": 11 Target string to be generated const string TARGET = "I love GeeksforGeeks": // Function to generate random numbers in given range int random_num(int start, int end) { int range = (end-start)+1: int random_int = starte(rand()‘érange): return random_int; } // Create random genes for mutation char mutated_genes() C int len = [Link](); int r= random_num(0, len-1); return GENES(r]: d // create chromosome or string of genes string create_gnome() { INTRODUCTION TO COMPUTATIONAL INTELLIGENCE int len = [Link](); string gnome = for(int i = Oziclensis+) gnome += mutated_genes(): return gnome: 3 // Class representing individual in population class Individual C publi string chromosome: int fitness: Tndividual(string chromosome): Individual mate(Endividual parent2): int cal_fitness(): ub Tndividual:Individual(string chromosome) { +this-»chromesome = chromosome; fitness = cal_fitness(); % // Perform mating and produce new offspring Individual Individual::mate(Individual par2) { // chromosome for offspring ‘string child_chromosome = ""; int len = [Link](): for(int i lensit+) { / vandom probability float p = random_num(0, 100)/100; // if prob is less than 0.45, insert gene // from parent 1 if(p < 0.45) child_chromosome += chromosome[iy: /1 if prob is between 0.45 ‘and 0.90, insert [Link] from parent 2 else if(p < 0.90) child_chromosome += [Link]'l // otherwise insert random gene(mutate), // for maintaining diversity else child_chromosome += mutated_genes(): // create new Individual(of spring) using // generated chromosome for of fspring return Individual(child_chromosome): i // Calculate fitness score, it is the number of // characters in string which differ from target 1 string, int Individual::cal_fitness() { int len = [Link](); int fitness = 0; for(int i = O;iclenzi++) { if(chromosome[i] Iz TARGETLI]) fitnesse+; 3 return fitness; z // Overloading < operator bool operator«(const Individual &indt, const Individual ind2) { return [Link] < [Link]; // Driver code int main() £ ‘srand((unsigned)(time(0))); // current generation ad INTRODUCTION TO COMPUTATIONAL iy, _INTRO Lg i int generation = vector«ndividual> population; bool found = false: // create initial population for(int i = 0;iPOPULATION_SIZE:ive) { ‘string gnome = create_gnome(): population push_back(Zndividual(gnome)) d while(! found) { // sort the population in increasing order of fitness score sort(population,begin(), population end(); // if the individual having lowest fitness score ig 7/0 4then we know that we have reached to the target 7/ and break the loop if(population{O} fitness <= 0) { found = true; break; // Otherwise generate new of springs for new generation vector new_generation; // Perform Elitism, that mean 10% of fittest population 1/ goes to the next generation int s = (10*POPULATION_SIZE)/100; for(int i= O;icsrie#) new_generation.push_back(population{i): // From 50% of fittest population, Individuls // will mate to produce offspring = (90*POPULATION_SIZE)/100;, for(int i = Orics:iee) S COMPUTATIONAL INTELLIGENCE int len = population size(): int e = rondom_num(0, 50): Individual parent! = population(r): = random_num(0, 50): Individual parent2 = population(e}: Individual of spring = [Link](parent2): rnew_generation.push_back(of fspring): } population = new_generation: coutee "Generation: * « generation «"\t"; couts« "String: "«« pepulation{O].chromasome «\ cout "Fitness: "«« population[0] fitness <<" generation: } coutes "Generation: " « generation « "\t": coutes "String; "<< population[O}.chromosome <"\t"; coutss "Fitness: "«< population{O],fitness < "\n"; } Lear Data Structures & Algorithms with GeeksforGeeks Output Generation: 1 Fitness: 18 Generation: 2. String: tOf Fitness: 18 Generation: 3 String: #IRWF9K Ifshw #OSk_ Fitness: 17 Generation: 4 String: -1Rq?3mHgk3Wo}3rek_ Fitness: 16 eneration: 5 String: -1Rq?8mHqk3Wo]3rek_ Fitness: 16 Generation: 6 String: AldW) #llksIw cVek) Fitness: 14 Generation: 7 String: A¥idW) #llksIw cVek) Fitness: 14 Generation: 8 String: (0x x06Rs=, 6Peek3 Fitness: 13 Generation: 29 String: Ilope Geeks#o, Geeks Fitness: 3 Generation: 30. String: IloMe GeeksfoBGeeks Fitness: 2 Generation: 31 String: Ilove Geeksfo0Geeks Fitness: 1 Generation: 32 String: Ilove Geeksfo0Geeks Fitness: 1 Generation: 33 String: Tove Geeksfo0Geeks Fitness: 1 Generation: 34 String: I love GeeksforGeeks Fitness: 0 Note: Every-time algorithm start with random strings, so output may differ. As we can see from the output, our algorithm sometimes stuck at 2 local optimum solution, this can be further improved by updating fitness score ‘alculation algorithm or by tweaking mutation and crossover operators. aay INTRODUCTION TO COMPUTATIONAL INTELLIGENCE Why use Genetic Algorithms? + They are Robust + Provide optimisation over large space state © Unlike traditional Al, they do not break on slight change in input or presence of noise. Application of Genetic Algorithms Genetic algorithms have many applications, some of them + Recurrent Neural Network. ‘+ Mutation testing, + Code breaking, ‘+ Filtering and signal processing. + Learning fuzzy rule base etc. Genetic Programming (GP): Similar to GAs, GP evolves computer programs or solutions to a problem using genetic operations like mutation and crossover. Fuzzy Systems: Refer Section 1.2. 1.4.2 Swarm Intelligence Particle Swarm Optimization (PSO) Inspired by the social behavior of birds and fish, PSO involves a population of particles that move through the search space to find optimal solutions. The process of finding optimal values for the specific parameters of a given system to fulfil all design requirements while considering the lowest possible cost is referred to as an optimization. Optimization problems can be found in all fields of science. Conventional Optimization Algorithms algorithms) have some limitations such as: ‘+ Single-based solutions * Converging to local optima + Unknown search space issues To overcome these limitations, many scholars and researchers have developed several metaheuristics to address complex/unsolved optimization problems. Example: Particle Swarm Optimization, Grey wolf optimization, Ant colony Optimization, Genetic Algorithms, Cuckoo search algorithm, etc. The Introduction to Particle Swarm Optimization (PSO) article explained the basics of stochastic optimization algorithms and explained the intuition behind Particle ‘Swarm Optimization (PSO). This article aims to deep dive into Particle Swarm Optimization (PSO), Inspiration of the Algorithm ‘+ Particle Swarm Optimization (PSO) is @ powerful meta~ heuristic optimization algorithm and inspired by swarm behavior observed in nature such as fish and bird schooling. = Ss (deterministic 4.12) __ INTRODUCTION TO COMPUTATION, fe Hyperparameters of the Algi © Number of particles (N) + Maximum numberof trations (may gy COMPUTATIONAL INTELLIGENCE PSO is a Simulation of a simplified social system ay orignal intent of 50 algorithm was to graphicaly simulate the graceful but unpredictable choreos! a of a bird flock : | % mie deen + Innature, any ofthe bird's observable vicinity is imited |, Cognition ofp to some range. However, having more than one is | * Co ntuence of swarm (C2) alows al he birds n swarm o be aware ofthelarger J : surface ofa fitness function 1y initialize Swarm popugy i fandomly Pulation * Let's mathematically model, the above-mentioned eter FN ag Principles to make the swarm find the global minima | x, (j= 1, 2,--» ee eh of a fitness function Step 2: Select hyp: Mathematical Model w, cL and ¢: Each particle in particle swarm optimization has a” | geoy 3: gor iter in range(max itn, associated position, velocity, fitness value. Each particle keeps track of the particle bestfitness,value particle. For i in range(N): # for each particle: Aredia pestis potion ane gual | PPEaetiener eye ea [Link] maintained. ‘swarm[i].velocity = Swarm w'swarmfilvelocity + <0 aia ret"(cwarm(iLbestPos- nara pasty r2*e2*(best_pos_swarm - swarmfi Posi Svarntt] jeseond parte b. Compute new position of ith particle un : 1 new velocity a mee swormtilpositon += swarmveleciy , 7 ©. Tf position is not in range {minx, mas the : i it Swarm] Last particle if swart{i] position « minx: Beal postion and finese Swarm swarrnfi} position = minx elif swarm[i] position > maxx Fig. 1.8: Data structure to store swarm population nL : ‘swarm[i].position = maxx gua] d. Update new best of this particle and new beste __ Swarmfi}.position ‘Swarm Le if swalnsensitive to scaling of design 2 [Link]] fitness < swarm{i] bestFitness: Swarm{i bestFitness = [Link] ‘Swarnlifiness Swarrn[i|.bestPos = swarm[il position [Swart besiPos 'f swarm[i] fitness < best_fitness_swerm = best_fitn = i] fitness aes gist Fitness_sworm = snern) '_Pos_swarm = swarm[i} position Fig. 1.9: Data structure to store ith particle of swarm end for : a5 Algorithm End -for oe : Parameters of Problem Step 4: Return b * Number of dimensions (d) Amariagilt; seein * Lower bound (minx) ages of PSO Upper bound (mao) * Insensitive to scaling of design variables a A ie i de COMPUTATIONAL INTELLIGENCE = Very few algorithm parameters. «Very efficient global search algorithm + Easily paralleled for concurrent processing Disadvantages of PSO Siow convergence in the refined search stage (Weak local search ability) 2.4.3 Ant Colony Optimization(ACO) ‘© Based on the foraging behavior of ants, ACO is used for solving combinatorial optimization problems. +The algorithmic world is beautiful with multifarious strategies and tools being developed round the clock to render to the need for high-performance ‘computing. In fact, when algorithms are inspired by natural laws, interesting results are observed. Evolutionary algorithms belong to such a class of algorithms. ‘+ These algorithms are designed so as to mimic certain behaviours as well as evolutionary traits of the human genome. Moreover, such algorithmic design is not only constrained to humans but can be inspired by the natural behaviour of certain animals as well. The basic aim of fabricating such methodologies is to provide realistic, relevant and yet some low-cost solutions to problems that are hitherto unsolvable by conventional + Different optimization techniques have thus evolved based on such evolutionary algorithms and thereby opened up the domain of metaheuristics. Metaheuristic has been derived from two Greek words, namely, Meta meaning one level above and heuriskein meaning to find. Algorithms such as the Particle Swarm Optimization (PSO) and Ant Colony Optimization (ACO) are examples of swarm intelligence and metaheuristics + The goal of swarm intelligence is to design intelligent multi-agent systems by taking inspiration from the collective behaviour of social insects such as ants, termites, bees, wasps and other animal societies such as flocks of birds or schools of fish. Background + Ant Colony Optimization technique is purely inspired from the foraging behaviour of ant colonies, first introduced by Marco Dorigo in the 1990s. Ants are eusocial insects that prefer community survival and sustaining rather than as individual species, They communicate with each other using sound, touch and pheromone. * Pheromones are organic chemical compounds Secreted by the ants that trigger a social response in VIKAS a3) INTRODUCTION TO COMPUTATIONAL INTELUGENCE members of same species. These are chemicals capable of acting like hormones outside the body of the secreting individual, to impact the behaviour of the receiving individuals, Since most ants live on the ‘ground, they use the soil surface to leave pheromone trails that may be followed (smelled) by other ants, © Ants live in community nests and the underlying principle of ACO is to observe the movement of the ants from their nests in order to search for food in the shortest possible path. Initially, ants start to move randomly in search of food around their nests. ‘+ This randomized search opens up multiple routes from the nest to the food source. Now, based on the quality ‘and quantity of the food, ants carry 2 portion of the food back with necessary pheromone concentration on its return path, + Depending on these pheromone trials, the probability of selection of a specific path by the following ants would be a guiding factor to the food source. Evidently, this probability is based on the concentration as well as the rate of evaporation of pheromone. It can also be observed that since the evaporation rate of pheromone is also a deciding factor, the length of each path can easily be accounted fo La LenS rood 2 ES Colony, Food Colony, (Ca Food Colony, Food In the Fig. 1.10, for simplicity, only two possible paths have been considered between the food source and the ant nest. The stages can be analyzed as follows: Stage 1: All ants are in their nest. There is no pheromone content in the environment. (For algorithmic design, residual pheromone amount can be considered without interfering with the probability) Stage 2: Ants begin their search with equal (0.5 each) probability along each path. Clearly, the curved path is the longer and hence the time taken by ants to reach food source is greater than the other. Stage 3: The ants through the shorter path reaches food source earlier. Now, evidently they face with 2 similar selection dilemma, but this time due to pheromone trail NMPUTATIONAL INTELLIGENCE fe charter path alteady available, probability of electian is high stage ute More ants return via the shorter path and itiy the pheromone concentrations Moreamt, due to the Concertration in the longer path reduces, decreasing the probability of selection of this path in further stages. Therefore, the whole colony gradually uses the shorter path in higher probabilities, So, path optimization is attained, Algorithmic Design + Pertaining to the above behaviour of the ants, an algorithmic design can now be developed. For impliity, @ single food source and single ant colony have been considered with just tuo paths of possible treversal. The whole ccenario can bbe realized through ‘weighted grephs where the ant colony and the food 2 (or nodes}; the paths serve as the edges and the pheromone values are the weights aszocieted with the edges. + Let the graph be G = (V, £) where V, E are the edges and the vertices of the graph. The vertices according to our consideration are V, (Source vertex - ant colony) ‘and Vz (Destination verter - Food source), The two edges are E: and E; with lengths La and Lo assigned to + Non, the sssociated pheromone values (indicative of their strength) can be assumed to be Ri and Ry for vertices E end E respectively. Thus for each ant, the starting probability of selection of path (between E: and 7) can be eraporation, ign -yoressed as follows Ba. 12 mek! Evidently, if Ps > Re, the probability of choosing Eis higher and vice-verse, Now, while returning through this shortest path say &, the pheromone value is updated for the corresponding path. The updation is done based on the length of the paths as well as the evaporation rate of pheromone. So, the update can be step-wise realized as follows: Jn accordance to path length - R Kk Be RAT In the above updation, i = 1, 2 and ‘K’ serves as a Parameter of the model. Moreover, the update is dependent on the length of the path. Shorter the path, higher the pheromone added In accordance to evaporation rate of pheromone ~ Rea-v th am INTRODUCTION TO COMPUTATIONAL ITELUcay, The parameter ‘v belongs 10 intenal (, 1) yy regulates the pheromone evaporation. Further,i= y's! « + Ateach iteration, all ants are placed at source vertex, (ant colony). Subsequently, ants move from V, 1g y, (food source) following step 1. Next, all ants cong their return trip and reinforce their chosen path basey fon step 2. Pseudocode Initialize necessary parameters and pheromone trial while not termination do: 7 Generate ant population: Calculate fitness values associated with each ant; g Find best solution through selection methods; Update pheromone tr “ end while < end procedure i +The pheromone update and the fitness calculations in '* the above pseudocode can be found through the step wise implementations mentioned above. a + Thus, the introduction of the ACO optimization technique has been established. The application of the ‘ACO can be extended to various problems such as the ¢ famous TSP (Travelling Salesman Problem), 7 1.4.4 Rough Set Theory 1 © This approach deals with uncertainty and approximation in data, It is particularly useful for ! feature selection and decision-making in the presence ! of incomplete or imprecise information 7 ‘+ The notion of Rough sets was introduced by Z Pawlak in his seminal paper of 1982 (Pawlak 1982). It is @ formal theory derived from fundamental research on logical properties of information systems. Rough set theory has been a methodology of database mining or knowledge discovery in relational databases, ‘In its abstract form, it is a new area of uncertainty mathematics closely related to fuzzy theory. We can use rough set approach to discover structural ‘elationship within imprecise and noisy data. Rough sets and fuzzy sets are complementary generalizations of classical sets. * The approximation spaces of rough set theory are sets with multiple memberships, while fuzzy sets are concerned with partial memberships. The rapid development of these two approaches provides a basis for "soft computing” initiated by Lotfi A. Zadeh. e Se ‘COMPUTATIONAL INTELLIGENCE 2 as) © Soft Computing includes along with rough sets, at least fuczy logic, neural networks, probabilistic reasoning, belief networks, learning, evolutionary computing and chaos theory. Basic Problems in Data Analysis Solved by Rough Set «Characterization of a set of objects in terms of attribute values, «Finding dependency between the attributes. «Reduction of superfluous attributes, + Finding the most significant attributes. machine ‘© Decision rule generation. Goals of Rough Set Theory The main goal of the rough set analysis is the induction of (learning) approximations of concepts. Rough sets constitute a sound basis for KDD. It offers mathematical tools to discover patterns hidden in data, It can be used for feature selection, feature extraction, data reduction, decision rule generation and pattern extraction (templates, association rules) etc. Identifies partial or total dependencies in data, eliminates redundant data, gives approach to null values, missing data, dynamic data and others (A) Information System In Rough Set, data model information is stored in a table 11. Each row (tuples) represents a fact or an object. Often the facts are not consistent with each other. In rough set terminology a data table is called an Information System. Thus, the information table represents input data, gathered from any domain. Table 1.2: Information Table ‘Attributes ‘Case | Temperature | Headache | Nausa | Cough | 4._| High Yes__|_No_| Yes 2_| Very high Yes__| Yes | No 3._[ High No__| No | No 4._[ High Yes__| Yes | Yes 5_| Normal Yes__| No | No Normal No | Yes | Yes Note: Rows of a table are called examples (objects, entities). Information system is a pair (U, A, U is a non-empty finite set of objects and A is a non-empty finite set of attributes. The elements of A are called conditional attributes. An Information table sometimes called decision fable when it contains decision attribute/attributes, Decision system is a pair of (U, A union {d)), where d is decision attribute (instead of one we can consider more decision attributes), INTRODUCTION TO COMPUTATIONAL INTELUGENCE Table 1.3, Attributed Decision Temporature | Headacho | Nausa | Cough | Flu. 1_| High ves | No | ves | Yes 2._| Very high ves__| Yes | No | No 3._| High No No | No |_No 4._| High Yes_| Yes | Yes | Yes 5._| Normat Yes__| No | No | No 6._| Normal no | Yes | Yes | Yes (8) Indiscerri TTables may contain many objects having the same features. ‘A way of reducing table size is to store only one representative object for every set of objects with same features. These objects are called indiscernible objects or tuples. With any P subset A there is an associated equivalence relation INDP) IND (P) = (ye 2] Vae Pa = ay) where, IND(P) is called indiscernibility of relation. Here, x and y are indiscernible from each other by attribute. Table 1.4 po [pe [es 0 ‘Object 1 2 03 Of 05 08 o|=|o}o}n|ufR| ofrm}o}s]—|—fs] {01, 02) (03,07, 010) {04} {05} (06) {08} (09) {ply {01, 02) (03, 07, O10} 104) (05) {06} {08} {09} (el) bdp = pe bp = am 216) INTRODUCTION TO COMPUTATIONAL NTELLIGty in a t sip =f (01,02) {03, 05, 07, 09, 010) eae Inabove example, IND({p1}) = ({01, 02}, {03, OS, 07, 09, 010}, {04, 06, O8)) O1 and 02 are characterized by the same valves of attribute pl and the value is 1. 03, 05, 07, 09, 010 are characterized by the same value of attribute pl and the value is 2,04, 06, O8 are characterized by the same value of attribute pl and the value is 0 The indiscemibltycelation is an equivalence relation. Sets that are indiscernible are called elementary sets Approximations It Is a formal approximation of a crisp set defined by its two approximations ~ Upper approximation and Lower approximation (i) Upper Approximation: Upper approximation is the set of objects which possibly belong tothe target set. UWE U/R:VAX #9) (i) Lower Approximation: Lower approximation is the set of objects that positively belong to the target set. RK = U{Ye U/R:YCX ‘+ Represents the positive region which contains the objects defintely belonging to the target set X. + Represents the negative region which contains the objects that can be definitely ruled out as a member of the target set X. ‘Represents the boundary region which contains the objects that may or may not belong to the target set X. + A set is said to be rough if its boundary region is, non-empty, otherwise the set is crisp. RK Set X A Possible upper approximation R(X) Possible lower approximation R(X) Possible boundary region Fig. 1.11 Let us discuss an example. The previous table is taken as information table, Example : Target set, X = ((01, 02, 03, 05, 08}, {04, 06, 07, 09, 010)) INd() = (01,021 (03,07,010, > £ (04, (051 (08), 08, 09,1 Px = (01,02, 05,08, ' {04, 06, 07, 09, O10) . PK = (01,02,03,07,010,05,08, * (03, 04, 06, 07, 09, 010, positive region = UPX = (01, 02, 05,08, 04,05, 0% Negative region = U-UPX = NULL Boundary region = (03, 07, 010) 1.4.5 _ Expert Systems Knowledge-Based Systems: Expert systems use human knowledge to solve specific problems. They consist of 3 knowledge base and an inference engine that mimicg human reasoning, 1.4.6 Hybrid Systems Integration of Techniques: Many real-world probleme benefit from combining multiple CI techniques. Hybrid systems involve the integration of different computational inteligence approaches to leverage their complementary strengths. 7_Machine Learning + Ensemble Learning: This involves combining multiple machine learning models to improve overall performance and robustness. ‘+ Reinforcement Learning: An area of machine learning where an agent learns to make decisions by interacting with an environment and receiving feedback in the form of rewards or penalties. 1.5 SYNERGIES OF COMPUTATIONAL INTELLIGENCE TECHNIQUES. 1. Hybridization + Combining Evolutionary Algorithms with Neural Networks: Evolutionary algorithms, such as genetic algorithms, can be used for optimizing the parameters of neural networks. This hybrid approach, often called Neuroevolution, can be especially effective in training complex neural network architectures. + Fuzzy Logic with Genetic Algorithms: Combining fuzzy logic for handling uncertainty with genetic algorithms for optimization can lead to robust systems that are capable of dealing with imprecise information. Example: Neuroevolution for Reinforcement Learning In the field of reinforcement leaming, genetic algorithms ‘can be used to evolve the architecture and parameters neural_networks that control agents in_an environment P = {pl p2, p3, p4, p5} (COMPUTATIONAL INTELLIGENCE This approach, known as neuroevolution, has been applied to tasks such as training game-playing agents and robotic control. The genetic algorithm optimizes the neural network's structure and weights over generations, leading to more effective policies, 2. Ensemble Methods + Ensemble of Neural Networks: Combining multiple neural networks into an ensemble can improve generalization and robustness. Techniques like bagging and boosting can be applied to neural networks to enhance overall performance. + Hybridizing Evolutionary Algorithms with Ensemble Learning: Integrating evolutionary algorithms with ensemble methods can lead to the evolution of diverse and effective individual models within the ensemble. Example: Neural Network Ensemble for Classification Image Instead of relying on a single neural network for image sification, an ensemble of neural networks can be treated. Each neural network in the ensemble may be rained with different initial conditions or subsets of the data, The final prediction is then determined by aggregating the individual predictions of each network. This ensemble approach often improves accuracy and ‘obustness compared to using a single neural network. 3. Fusion of Data and Knowledge 19 Fuzzy Systems with Expert Systems: Combining ‘uzzy systems for handling uncertainty with expert systems that encode human knowledge can create powerful decision support systems capable of dealing with imprecise information and complex decision-making scenarios. Example: Medical Diagnosis System in a medical diagnosis system, fuzzy logic can be used to sandle uncertainty in patient data (e.g, imprecise test sults). Expert knowledge, encoded in an expert system, rovides rules for diagnosing medical conditions. The “usion of fuzzy logic and expert systems allows the system ‘0 make informed and interpretable diagnoses, even when dealing with incomplete or uncertain information, 4. Reinforcement Learning with Evolutionary Algorithms Combi ing Reinforcement Learning and Genetic Algorithms: Evolutionary algorithms can be used to optimize the parameters of reinforcement learning agents, helping to speed up the learning process and enhance the Performance of the agent in complex environments. Example: Evolving Neural Network Controllers for Robotics aan INTRODUCTION TO COMPUTATIONAL INTELUIGENCE Genetic algorithms can be employed to optimize the parameters of neural networks that act as controllers for robotic systems. In a reinforcement learning setting, the genetic algorithm evolves neural network structures and weights to produce controllers that enable robots to learn and adapt to different tasks in a dynamic environment. 5. Combining Particle Swarm Optimization with Ant Colony Optimization: Integrating different swarm intelligence techniques can lead to improved optimization performance. For example, combining the global exploration capability of particle swarm optimization with the local search ability of ant colony optimization can enhance the convergence speed and quality of solutions. Example: Swarm-Enhanced Particle Swarm Optimization Integrating Particle Swarm Optimization (PSO) with Ant Colony Optimization (ACO) can be beneficial for optimization problems, PSO provides global exploration capabilities, while ACO contributes local search abilities. This hybrid approach can improve convergence speed and solution quality in tasks such as optimization of complex functions or parameter tuning in machine learning models. 6. Neuro-Fuzzy Systems Integrating Neural Networks with Fuzzy Logic: Neuro- fuzzy systems combine the learning capabilities of neural networks with the interpretability of fuzzy logic. This ‘combination is useful in applications where both learning from data and handling uncertainty are crucial Example: Adaptive Traffic Control System In a traffic control system, a neuro-fuzzy approach might be used to optimize traffic signal timings. Neural networks can learn from real-time traffic data, while fuzzy logic can handle the uncertainty and imprecision inherent in traffic patterns. The integrated neuro-fuzzy system adapts traffic signal timings to optimize traffic flow based on current conditions, 7. Hybridization for Real-world Applications Combining CI with Conventional Algorithms: Integrating CI techniques with traditional optimization or control algorithms can lead to hybrid systems that leverage the strengths of both approaches. This is particularly beneficial in real-world applications. where the computational cost of pure CI methods may be a concern. Example: Hybrid Genetic Algorithm for Job Scheduling In job scheduling problems, a hybrid approach could involve combining a genetic algorithm with traditional ‘scheduling algorithms. The genetic algorithm optimizes the overall schedule, while the conventional algorithm handles specific constraints or aspects of the problem that are well- suited to deterministic approaches. COMPUTATIONAL INTELLIGENCE 8, Feature Selection and Dimensionality Reduction Genetic Algorithms with Machine Learning Models: Genetic algorithms can be employed for feature selection, helping machine learning models focus on the most relevant features. This can enhance model interpretability, reduce overfitting and improve generalization Example: Genetic Algorithm for Feature Selection in Cancer Diagnosis In the context of cancer diagnosis using gene expression data, a genetic algorithm can be applied to select the most relevant genes as features for a machine learning model. This enhances the model's interpretability and reduces the dimensionality of the input space, potentially leading to improved classification accuracy and the identification of ctitical biomarkers. [1.6 APPLICATIONS OF COMPUTATIONAL as) INTELLIGENCE + Tackling Complex Problems: Computational inteligence offers solutions to complex problems that traditional algorithmic approaches may struggle to solve. Examples of its applications span various domain. + Natural Language Processing: With computational intelligence techniques, machines can understand, interpret and generate human language in written or spoken forms. Applications include sentiment analysis, machine translation and chatbots. * Image and Video Processing: Using techniques like neural networks, computational intelligence enables image and video processing for tasks such as object detection, face recognition and video analytics. ‘+ Roboties and Automation: In robotics, computational intelligence is used to develop robots with increased autonomy and adaptability. Applications range from self-driving cars and drones to robotic arms. in manufacturing and medical robotics. * Decision Support Systems: Computational intelligence supports decision-making in complex environments. Examples include stock market prediction, risk assessment in finance and recommendation systems for e-commerce or content platforms. * Bioinformatics and Medicine: In bioinformatics and medicine, computational intelligence techniques help analyze vast amounts of data, such as genomics data ‘or medical images, to find patterns, perform diagnoses and predict patient outcomes. Sv INTRODUCTION TO COMPUTATIONAL INTELUGENey 7 GRAND CHALLENGES OF COMPUTATIONAL INTELLIGENCE Computational Intelligence (Cl) faces several grang challenges that researchers and practitioners are active working to address. These challenges encompass varios fspects of Cl. including algorithmic advancements, rel. World applications, ethical considerations and societa| impacts, Here are some of the grand challenges in Computational intelligence: + Explainability and Interpretability: Developing models that are not only accurate but also transparent and interpretable. Understanding and explaining the decision-making process of complex models, especialy in critical applications like healthcare and finance, is crucial for gaining trust and acceptance. + Transferability and Generalization: Enhancing the ability of CI algorithms to transfer knowledge across different domains and generalize well to unseen data Improving the robustness and adaptability of models to ensure their performance in diverse and dynamic environments. ‘+ Scalability and Efficiency: Scaling Ci algorithms to handle large datasets and complex problems efficiently. Developing methods that can leverave paral! and distributed computing to address the computational demands of modern applications. + Autonomous Learning and Adaptation: Enabling @ systems to autonomously learn and adapt to changing environments without significant human intervention This includes developing algorithms that cen continuously learn from new data and update their models to maintain relevance over time. + Ethical and Fair AL Addressing ethical concerns felated to bias, faimess and accountability in @ systems. Developing methods to ensure that CI models are fair and unbiased across diverse demographic groups and establishing guidelines for responsible and ethical Al development. * Real-world Robotic Applications: Advancing Cl fot robotics to enable robots to operate in unstructured ‘and dynamic environments. Overcoming challenges related to sensorimotor integration, path planning and human-robot interaction to enhance the practi deployment of robotic systems in fields such & healthcare, manufacturing and disaster response. * Security and Trustworthiness: Ensuring the secuttY and. trustworthiness of Cl models and systems ¢ ies such as adversarial attecl’ Addressing vulnerabi and developing ‘on machine learning models techniques to enhance the robustness and resilience Ci systems in the face of security threats. h © _—_a as) INTRODUCTION TO COMPUTATIONAL INTELUGENCE COMPUTATIONAL INTELLIGENCE 7 © Human-Centric Computing: — Integrating — Cl technologies with human-centered design principles to create systems that augment human capabilities and nnance user experience. Designing CI systems that sre intuitive, user-friendly and adaptable to individual user needs. = Cognitive Computing and Artificial General Intelligence (AGI): Advancing Cl towards achieving artificial general intelligence, which developing systems capable of performing any intellectual task that a human being can, This encompasses higher-level cognitive abilities such as reasoning, problem-solving and abstract thinking. involves «Societal and Ethical Impacts: Understanding and addressing the broader societal impacts of Cl, including issues related to job displacement, economic inequality and the ethical implications of widespread adoption. Developing policies and frameworks to ensure responsible and equitable deployment of CI technologies. «© Interdisciplinary Collaboration: Promoting collaboration between CI researchers and experts from diverse fields such as neuroscience, psychology and social sciences. Integrating insights from different disciplines to advance our understanding of, intelligence and improve the design of more human- like CI systems. ————___________ (CASE STUDY : STUDY OF INTELLIGENT WASTE CLASSIFICATION SYSTEM USING COMPUTATIONAL INTELLIGENCE. Introduction Waste management is a critical challenge in_urban environments and the need for efficient waste classification has become increasingly important for sustainable practices. In this case study, we explore the implementation of an Intelligent Waste Classification System (IWCS) using Computational Intelligence techniques to enhance waste sorting and recycling processes. Objective The primary objective of the Intelligent Waste Classification System is to automate the sorting of waste into various categories such as recyclables, non-recyclables organic and hrazatdous materials. The system aims to improve the efficiency of waste management processes, reduce Contamination and promote recy Components of the Intelligent Waste Classification System 1. Sensor Network: Deploy a network of sensors to 5. collect data on waste composition, weight and other Televant parameters. These sensors are strategically placed in waste bins or collection points, Data Acquisition and Preprocessing: Computational Inteligence algorithms to process rav data collected from sensors. Apply techniques such a3 data filtering and normalization to enhance the quality Utilize of input data, Feature Extraction: Employ machine learning algorithms for feature extraction to identify key characteristics of different types of waste. Features may include color, texture, shape and weight. Classification Model: Train a classification model Using techniques such as Support Vector Machines (SVWMh, Neural Networks or Decision Trees. The model learns from labeled data to accurately classify waste items into predefined categories. Feedback Loop: Implement 2 feedback loop. ‘mechanism where the system continuously learns and adapts based on new data. This ensures the model evolves over time, improving its accuracy and adaptability to changing waste compositions. Implementation Imagine 2 city deploying the Intelligent Waste Classification System in a pilot project: 5. Sensor Deployment: Install sensors in waste bins across different neighborhoods. These sensors include cameras, weight sensors and environmental sensors to capture data. Data Collection: Sensors continuously collect data on the waste being disposed of. The data is transmitted to a central processing unit for further analysis. Data Processing and Feature Extraction: Computational Intelligence algorithms process the data, extracting relevant features such as color distribution, texture and weight. The system identifies patterns associated with recyclables organics and other waste categories. Model Training: Train a machine learning model using historical data labeled by waste category. The model learns to recognize patterns and associations between features and waste types. Real-time Classification: Implement the _ trained model in the waste bins’ processing units to classify incoming waste in real-time. The system categorizes the waste into recyclables, non-recyclables organics and hazardous materials (COMPUTATIONAL INTELLIGENCE ee 6. Adaptive Learning: Integrate an adaptive learning mechanism to continually update the model based on new waste data, This ensures the system remains effective even as waste compositions change over time, Results: The implementation of the Intelligent Waste Classification System yields several Advantages. + Improved Recycling Rates: By accurately sorting waste at the source, the system enhances recycling rates and reduces the contamination of recyclable materials, * Operational Efficiency: Waste management processes become more efficient with automated classification, reducing the need for manual sorting and streamlining collection and processing. + Environmental Impact: The system contributes to environmental sustainability by promoting proper waste disposal practices, leading to reduced landfill usage and a decrease in environmental pollution. Challenges and Future Developments * Adaptability to Changing Waste Streams: The system must continuously adapt to changes in waste compositions due to seasonal variations or shifts in consumption patterns. + Scalability: As the system proves successful in a pilot project, scalability to cover larger urban areas is a consideration for widespread adoption. ‘+ Integration with Smart Cities: Integration with broader smart city initiatives can enhance overall urban sustainability and efficiency. EXERCISE 1. Define Computational Intelligence and highlight its key objectives. 2. Explain the significance of adaptive systems in the context of Computational intelligence. 3. Provide examples of real-world problems where Computational Intelligence techniques are applicable. 4. Compare and contrast the three main paradigms of Computational Intelligence: Neural Networks, Evolutionary Computation and Fuzzy Systems, 2.20) 10, 1 12, 3 14, 15, 16. v7. 18. 19, 20, 21. INTRODUCTION 70 COMPUTATIONALINTELLIGEi, Discuss a specific application scenario for gq, paradigm, highlighting its strengths and weaknesser How do these paradigms complement each other ip solving complex problems? Outline the fundamental differences between Artificial Intelligence (Al) and Computation Intelligence (CD. . Provide examples of Al applications that may not fa under the scope of Computational Intelligence. Explain why a distinction between AI and Ct is important in problem-solving approaches. Describe the main approaches within each paradigm of Computational Intelligence. Discuss how rule-based systems are used in Fuzzy Systems and their role in decision-making, Highlight the role of genetic operators in Evolutionary Computation and their impact on optimization, Explore scenarios where combining Neural Networks, Evolutionary Computation and Fuzzy Systems provides a synergistic advantage. How can hybrid models leverage the strengths of different Computational Intelligence paradigms? Provide an example where the synergy of techniques leads to improved problem-solving, Choose a specific industry (e.g,, healthcare, finance or manufacturing) and discuss how Computational Intelligence techniques are applied in solving problems within that industry. Explain the role of Computational Intelligence in image recognition and natural language processing, Explore emerging applications of Computational Intelligence in current technological trends. Identify and discuss the grand challenges faced by Computational Intelligence in solving complex problems. How does the scalability of Computational Intelligence techniques pose a challenge in large- scale applications? Explore ethical considerations related to the use 0 Computational Intelligence in decision-making Processes. eR UNIT - 1) FUZZY LOGIC ——nnnnnneneeee 2.1 INTRODUCTION TO FUZZY SET 2.1 Fuzzy Numbers “= Fuzzy numbers are a fundamental concept in the field of fuzzy logic, which was introduced by Lotfi A. Zadeh in the 1960s as @ way to represent and manipulate uncertainty and vagueness in data and decision- making, * Fuzzy numbers extend the traditional notion of real numbers by allowing for degrees of membership or possibility, making them a valuable tool for modeling and analyzing imprecise and uncertain information. «At their core, fuzzy numbers provide a framework to represent and quantify uncertainty in a more flexible manner than traditional crisp (or non-fuzzy) numbers. ‘A fuzzy number is defined by @ membership function that assigns a degree of membership to each real number within a certain range. + This membership function describes the extent to which a given value belongs to the fuzzy number. The membership function typically takes on values between 0 and 1, where 0 indicates no membership, 1 indicates full membership and values in between reflect partial membership. |The key elements of a fuzzy number are its support core and shape. The support represents the range of values for which the fuzzy number has non-zero membership. Within this support, the core is the region where the membership value is equal to 1, indicating maximum membership. The shape of the membership function characterizes how the degree of membership varies across the support, defining the specific characteristics of the fuzzy number. ‘One common type of fuzzy number is fuzzy number, which is characterized by a triangular- shaped membership function. Triangular fuzzy numbers are widely used because of their simplicity and ease of interpretation. They are defined by three parameters: a lower bound (@) an upper bound (b) and a peak value (c) where gph Gugee Ine members function of a triangular the triangular en fuzzy number is 0 for values outside the range (2, b} and it increases linearly from 0 to 1 as the value moves from a to c and then decreases linearly from 1 to 0 as the value moves from cto b. Fuzzy numbers can also take on other shapes, such as trapezoidal, Gaussian or any custom shape that adequately represents the uncertainty in a given context. These different shapes allow for more precise modeling of various types of uncertainty. ‘One of the significant advantages of fuzzy numbers is their ability to represent imprecision and uncertainty in a natural and intuitive way. Traditional real numbers are not well-suited to handle uncertain information, 25 they assume perfect precision and do not account for the inherent vagueness in many real-world situations. Fuzzy numbers, on the other hand, exce! at capturing and managing uncertainty, making them valuable in fields such as decision-making, control systems and risk analysis. In decision-making, fuzzy numbers are used to model vague or ambiguous information. For example, when evaluating the performance of a product based on customer feedback, the satisfaction level may be represented as a fuzzy number. This allows decision-makers to account for the fact that some customers are very satisfied, some are moderately satisfied and others are only slightly satisfied, rather than forcing a binary “satisfied” or “not satisfied” categorization. In control systems, Fuzzy numbers are employed to design controllers that can handle imprecise inputs ‘and provide robust performance in uncertain environments, Fuzzy logic controllers use fuzzy numbers to define rules that guide the control actions based on the current system state, allowing for more flexible and adaptive control. In tisk analysis, fuzzy numbers are used to model uncertain parameters. in probabilistic models. For example, in financial risk assessment, the expected feturn on an investment may be represented 2s 2 fU22Y vrmber to account for the uncertainty In future market conditions. COMPUTATIONAL INTELLIGENCE * Fuzzy numbers also find applications in optimization problems, where traditional crisp numbers may not accurately represent the objectives and constraints Fuzzy optimization techniques aim to find solutions that satisfy fuzzy objectives and constraints, taking into account the degrees of membership associated fuzzy numbers. + In summary, fuzzy numbers are @ powerful tool for hhandiing uncertainty and vagueness in various domains. They provide a flexible and intuitive way to represent and work with imprecise data, making therm 2 valuable asset in decision-making, control systems, risk analysis and optimization. © The concept of fuzzy numbers has. significantly contributed to the development of fuzzy logic and its practical applications in addressing real-world problems where uncertainty is a prevalent and unavoidable aspect of the data and decision-making process, To learn about classical and fuzzy set theory, firstly you have to know about what is set. Set: A set is a term, which is a collection of unordered or ordered elements. Following are the various examples of a set: © Aset of all-natural numbers * Asset of students in a class. + Asset of all cities in a state. © Aset of upper-case letters of the alphabet Types of Set There are following various categories of set: © Finite + Empty © Infinite + Proper = Universal + Subset Equivalent Set igleton . Disjoint Set (A) Classical Set It is a type of set which collects the distinct objects in @ group. The sets with the crisp boundaries are classical sets. In any set, each single entity is called an element or member of that set. Mathematical Representation of Sets Any set can be easily denoted in the following two Jifferent ways: 22) ww - FUZZY tog 1. Roaster Form: this ic alvo called as a tabain tayy, this forrn, the vet ted in the following may Set_name = (element, element, eheHNeENN, on, element Ny The elernents in the set are enclosed within yy brachats and separated by the coronas Folloviing are the two ezamples which describes thy set Example}: Set of Natural Numbers, N= (0,2,2,4,5,6, ay Set of Prine Nurnbers less than 50, | [ X= (3, 5,7,21,13,17, 19,22, 29,31, 37, 41.43.47) | 2. Set Builder Form: Set Builder form defines a set with the cornmon properties of an element in a set. In this form, the set is represented in the following way: A= bap) The following evample describes the set in the builder Example 2: form The set (2, 4,6, 8, 10, 12, 14, 16, 18} is written as, B = (4221 < 20 and (7%2) = 0} (B) Fuzzy Set Theory (Continuous) + Fuzzy set theory, a foundational concept in fuzzy logic, extends classical set theory by allowing for the representation of imprecise and uncertain information + Unlike traditional sets that categorize elements as either fully belonging or not belonging to a set, fuzzy sets introduce the notion of partial membership. This extension of classical set theory provides @ powerful framework for handling ambiguity and vagueness in various applications. ‘+ In the context of continuous fuzzy sets, the primary distinction lies in how membership values are assigned to elements within a universe of discourse. Unlike discrete fuzzy sets, which categorize elements discretely, continuous fuzzy sets operate ove! continuous domains, making them particularly valuable for modeling and analyzing phenomena that exhibit # smooth transition of membership degrees. + A continuous fuzzy set is defined by a membership function that maps each element in the universe of discourse to a value between 0 and 1, representing the degree of membership. This membership function allows for @ more fine-grained representation uncertainty compared to discrete fuzzy sets. TONAL INTELLIGENCE “oMPUTATIONA —— he membership values can vary continuously, indicating the degree to which an element belongs to the fuzzy set. The integral of the membership function over the entire universe of discourse is equal to 1, ensuring that the membership values collectively cover the entire domain, ‘he defining characteristic of continuous fuzzy sets is their ability to represent gradual transitions and nuanced degrees of membership. This makes them suitable for applications where elements may exhibit varying levels of association with a set. + For example, in weather forecasting, @ continuous fuzzy set can be used to represent the likelihood of rainfall over a specific range of precipitation values, with the membership function smoothly transitioning from 0 to 1 as the amount of precipitation increases, ‘+ Continuous fuzzy sets find applications in a wide range of fields, including control systems, pattern recognition, decision-making and risk assessment. In control systems, they enable the design of controllers that can smoothly adapt to changing conditions, ensuring robust performance in dynamic environments. + In pattern recognition, continuous fuzzy sets are used to model complex patterns with varying degrees of membership to different classes, enhancing the accuracy of classification tasks. In decision-making, they allow for the representation of imprecise preferences and criteria, enabling more realistic and flexible decision models. In risk assessment, continuous fuzzy sets are employed to model uncertain parameters with a high degree of granularity, enhancing the precision of risk estimates. * In summary, continuous fuzzy set theory extends classical set theory by introducing the concept of Partial membership over a continuous domain. * Tt provides a versatile framework for representing and handling uncertainty in a smooth and nuanced manner, making it a valuable tool in various applications where imprecise and uncertain information is prevalent. © Fuzzy Set Theory (Discrete) Fuzzy set theory, a fundamental concept in fuzzy logic, ‘evolutionizes classical set theory by introducing the foncept of partial membership, allowing for a more ‘leible and realistic representation of uncertainty and —Haveness. 23) FUZZY LOGIC ‘While continuous fuzzy sets operate over continuous domains, discrete fuzzy sets work with discrete ‘elements or categories, making them suitable for applications where discrete categorization is more appropriate. In a discrete fuzzy set, consists of discrete elements and each element is assigned 2 meribership value between 0 and 1 indicating the degree to which it belongs to the fuzzy set. Unlike classical sets, which categorize elements as either entirely inside or outside a set, discrete fuzzy sets allow for elements to have partial membership, reflecting the inherent imprecision in many real-world situations. One common example of discrete fuzzy sets is linguistic variables. Linguistic variables are used to represent qualitative or subjective concepts, such as “low,” "medium," and "high" temperature. These linguistic variables are associated with fuzzy sets that assign membership values to different temperature values. For instance, the fuzzy set "low temperature” may have a membership function that assigns high membership values to temperatures below a certain threshold and gradually decreases membership values. as temperatures ise, reflecting the gradual transition from "low" to “not low" temperatures. Discrete fuzzy sets find applications in various domains, including expert systems, decision-making and control systems. In expert systems, they are used to represent imprecise knowledge and expertise, allowing for more nuanced reasoning and decision- making. In decision-making, discrete fuzzy sets enable the modeling of imprecise preferences and criter facilitating more realistic and flexible decision models. In control systems, they are employed to deal with imprecise inputs and outputs, making control systems more adaptable and robust in uncertain environments, One of the key advantages of discrete fuzzy sets is their simplicity and ease of interpretation. They provide 2 natural way to capture the qualitative aspects of data and knowledge, making them accessible to experts ‘and non-experts alike. Discrete fuzzy sets also allow for the development of rule-based systems that can handle uncertainty in a transparent and intuitive manner. the universe of discourse COMPUTATIONAL INTELLIGENCE 24) 2.2 MEMBERSHIP FUNCTION 2.2.1 Introduction of Membership Fun ® The membership function is a function which represents the graph of fuzzy sets and allows users to ‘quantify the linguistic term. It is a graph which is used for mapping each element of x to the value between Oand 1. + This also known as characterises function. This function of Membership was introduced in the first papers of fuzzy set by Zadeh, For the Fuzzy set 8, the membership function for X is defined as: BX = (01, In this function X, each element of set B is, mapped to the value between 0 and 1. This is called a degree of membership or membership value. function is indicator of ‘+ Fuzzy set membership functions play a crucial role in fuzzy logic systems by defining the degree to which an element belongs to a fuzzy set. Unlike classical set theory, which assigns a binary membership (either 0 or 1), fuzzy sets allow for degrees of membership ranging between 0 and 1. These membership functions are fundamental in representing and handling uncertainty ‘and imprecision in various applications. 2.2.2 Definition of Fuzzy Set Membership Function A fuzzy set A is characterized by a membership function denoted as wA(x), where x is an element of the universal set. The function wA(s) assigns a degree of membership between 0 and 1 to x, indicating the extent to which x belongs to the fuzzy set A. The higher the membership degree, the stronger the association of x with the fuzzy set. ‘Types of Membership Functions There are several types of membership functions, each suited for different types of fuzzy sets: () Triangular Membership Function: This is one of the simplest forms of membership functions, characterized by 2 triangular shape. It has three parameters: the left foot, the peak and the right foot. The degree of membership increases linearly from the left foot to the peak and decreases linearly from the peak to the right foot. Trapezoidal Membership Function: Similar to the ‘triangular function, the trapezoidal _ membership function has four parameters, defining the left foot, eft slope, right slope and right foot. It allows for a flatter top, offering more flexibility in representing fuzzy sets. FURY Log ii) Gaussian Membership Function: The Gay, used when the distribution of membership is expert, to follow a normal distribution. It is detineg parameters such as the mean and standard devistign ” Sigmoidal Membership Function: Sigmoig, functions have an S-shaped curve and are useq’ situations where there is 2 gradual transition betwee membership and non-membership. They ay, ‘commonly employed in control systems. ‘ 4 Applications of Fuzzy Set Membership Functions = Control Systems: Fuzzy logic controllers use membership functions to fuzzify input variables, converting crisp values into fuzzy sets. These membership functions capture the linguistic variables ‘and facilitate the application of fuzzy rules in the control process. + Pattern Recognition: Fuzzy set membership functions ate employed in pattern recognition systems where the boundaries between classes are not well-defined. They allow for a gradual transition from one class to another based on the degree of membership. + Decision Support Systems: In decision-making processes where uncertainty is inherent, fuzzy set membership functions help represent and model imprecise information. This is particularly useful in fields such as finance and risk assessment. Shape and Tuning The shape of the membership function directly influences the behavior of the fuzzy logic system. The choice of specific membership function depends on the characteristics of the system being modeled. Tuning the parameters of the membership functions allows for customization to match the specific requirements of the application. Conclusion Fuzzy set membership functions are a fundamental concept in fuzzy logic, enabling the representation of Uncertainty and imprecision in a wide range of applications. The choice of membership functions and the parameters is a critical aspect of designing effective fuzz) logic systems, allowing them to capture the complexity and nuances of real-world systems. The flexibility provided by membership functions contributes to the success of uz logic in addressing problems where classical methods m=! fall short due to uncertainty sraTIONAL INTELLIGENCE Fu ‘Operation on Classical Set 2 jallowi nthe classical sets 4. Union Operation 3. Intersection Operation jag are the various operations which are performed 3, Difference Operation 4, Complement Operation 4. Union Operation ‘This operation is denoted by (A U B). AU Bis the set of two different sets A and 8, This operation combines all the elements from both the sets and make a new set. It is also called a Logical OR operation. Itcan be described as: AUB = (ke AORXEB} those elements which exist i Example: Set A = (10,11, 12,13}, SetB = (11, 12,13, 14,15), then AUB = (10,11, 12, 13, 14, 15) 2. Intersection Operation This operation is denoted by (AN 8). A Bis the set of those elements which are common in both set A and 8. Itis also called a Logical OR operation Itcan be described as: ANB = (xe AAND xe B) Example: SetA = (10,11, 12,13), Set = (11,1214) then ANB = (11, 12) 3. Difference Operation This operation is denoted by (A ~ B). A-B is the set of only those elements which exist only in set A but not in set Itean be described as: A-B = fk AANDxEB) 4, Complement Operation This operation is denoted by (A). It is applied on a single set. A’ is the set of elements which do not exist inset A Itcan be described as: W = bee A. “VIRAS— 2s) FUZZY LOGIC 2.3.2 _ Properties of Classical Set There are following various properties which play an essential role for finding the solution of a fuzzy logic problem, 1. Commutative Property This property provides the following two states which are obtained by two finite sets A and 6: AuB = BUA ANB = BNA 2. Associative Property This property also provides the following two states but these are obtained by three different finite sets A, Band C: uu = (AUB)UC ANNO = (ANB)NC 3, Idempotency Property ‘This property also provides the following two states but for a single finite set A: AUA=A AMA=A 4. A’bsorption Property This property also provides the following two states for any two finite sets A and B: AU(ANB) =A AN(AUB) =A 5. Distributive Property ‘This property also provides the following two states for any three finite sets A, B and C: AUBNG = AUBNAVO AN BUC = (ANBUANG 6. Identity Property This property provides the following four states for any finite set A and Universal set X: AUG=A ANX =A Ane = 6 AUX =X 7. Transitive Property ‘This property provides the following state for the finite sets A, B and C: ASBSC thenASC — TATIONALINTELLIGENCE come Property avatation This property provides fatloting state for any finke set N ASA shaw De Morge the following tules for providing the This law giv contiadiction and lautologles: Anil = AUB Avi = Ant [2a Fuzzy SET CHARACTERISTICS 1. Membership Function: The — most — defining hoiactoustc ofa fuzzy set is its membership function. ats, which assign elements a binary Unlike cvisp membership value of either 0 or 1, fuzzy sets use membership functions to express the degree to which an element belongs to the set. The membership function, denoted as AO), maps each element x from the universal set toa value between O and 1, representing the degree of membership of x in the fuzzy set A, 2. Degree of Fuzziness: Fuzzy sets introduce the concept ‘of “luzziness,” which quantifies the extent to. which possess intermediate degrees of membership. The degree of fuzziness reflects the inherent uncertainty or lack of precision in the izzy sets are well-suited for boundaries between definition of a set situations categories are vague or subjective. 3, Overlap and Intersection: Unlike classical sets, fuzzy sets can overlap. The intersection of two fuzzy sets A and B is determined by taking the minimum of their respective membership values for each element. This allows for the representation of elements that simultaneously belong to multiple fuzzy sets, accommodating complex relationships in real-world modeling wher scenarios. A. Fuzzy Set Operations: Fuzzy sets employ a set of operations, including union, intersection and complement, which are extensions of classical set ‘operations adapted for degrees of membership. The union of two fuzzy sets, denoted as AUB, combines their membership values using the _ maximum operation, while the intersection, denoted as ANB, uses the minimum operation. Complement operations are used to express the degree to which an element does VWI REgNG 10 a fuzzy set 2.6) acesaeee Granularity and Resoluti axjustment of granularity oF resolution in te det of sets. Granularity refers to the degree of deta,” precision in spectying the boundaries ofa furry, finer granularity provides a more detaled disiny, between membership degrees, while _ granularity results in broader distinctions 6 Linguistic Variables: Fuzzy sets often linguistic variables to represent qualitative terns xu igh “low.” “warm, or "cold" Linguistic varie, coang, as enable the expression of human-like reasoning ang decision-making, making fuzzy logic suitable fa, modeling systems where imprecise language is used r, describe relationships and conditions. utilize continuous membership functions, allowing for a smooth transition between membership degrees This continuous nature enhances the modeling capabilities of fuzzy sets, especially in applications where gradual changes and transitions are important 8, Applications: Fuzzy sets find applications in diverse fields, including control systems, artificial intelligence, decision support systems and pattern recognition. Their ability to handle uncertainty and imprecision makes them well-suited for modeling real-world systems where precise boundaries and definitions are challenging, Conclusion: The characteristics of fuzzy sets, such as membership functions, degree of fuzziness and linguistic variables, make them 2 powerful tool for representing and reasoning with uncertain and imprecise information. The flexibility provided by fuzzy sets has led to their widespread adoption in various disciplines, offering framework for addressing complex and valuable ambiguous problems, |2.5_FUZZINESS AND PROBABILITY 2.5.1 Introduction of Fuzziness and Probability Fuzziness and probability are two distinct concepts, but they share some similarities in their attempts to handle Uncertainty. Both concepts arise from the recognition that in many real-world situations, information may be imprecise, incomplete or subject to variation. 1. Fuzziness ‘+ Fuzziness, in the context of fuzzy logic, refers to the idea that the boundaries between categories or classe are not sharply defined. _

Das könnte Ihnen auch gefallen