A Unit of Soundarya Educational Trust (R)
SOUNDARYA CENTRAL SCHOOL
Affiliated to CBSE – New Delhi
Artificial Intelligence (417)-Class: X- NOTES
Unit-2: Advanced Concepts of Modelling in AI
Reflection Time:
Answer the following questions:
Q1. Give difference between rule based and learning based AI models.
The main differences between rule-based and learning-based AI models are:
Rule-Based AI Models: The developer defines the relationships or patterns in data, and the machine
explicitly follows these rules. Learning is static, meaning the model will not learn from mistakes or
improvise based on feedback if tested on new or different data. An example is a rule-based chatbot.
Learning-Based AI Models: The computer learns by analyzing examples or receiving feedback,
finding patterns or rules on its own. These models are adaptive to changes in data, modifying
themselves to handle exceptions. This approach includes machine learning and deep learning. An
example is a learning-based spam email filter that continuously adapts and improves.
Q2. What is supervised, unsupervised and reinforcement learning? Explain with examples.
Supervised learning involves training AI models using labelled data, where the desired output is
known, similar to a teacher guiding a student with solved examples. An example is an email spam
filter that learns from emails explicitly marked as "spam" or "legitimate" to classify new emails.
Unsupervised learning deals with unlabelled datasets, where the AI independently finds patterns,
relationships, and trends in the data without prior knowledge. An example is an OTT platform
recommending content based on a user's watch history, identifying preferences without explicit
categories.
Reinforcement learning is a trial-and-error approach where the AI learns by making decisions to
maximize a reward metric, receiving positive or negative feedback. An example is an autonomous
vehicle learning to navigate by receiving rewards for safe driving and penalties for violations.
Q3. What is clustering and how is it different from classification?
Clustering is an unsupervised learning model where the machine identifies relationships, patterns, and
trends in unlabeled data without prior information. Its purpose is to group data based on inherent
similarities, like grouping dog images by color or size, or customers by purchase patterns, even without
initial labels.
Clustering finds similarities in unlabeled data to group objects into clusters, discovering
hidden patterns rather than predicting predefined ones.
Classification is a supervised learning model where the machine is trained using labeled data to classify
data into predefined categories. It predicts a specific class label as an output. Examples include
classifying emails as spam or not spam, or predicting weather as "hot" or "cold" based on historical
labeled data.
Classification uses predefined classes and labeled data to assign objects to known
categories
Q4. Explain neural networks. Also give functions of three layers of neural networks
Neural Networks are a type of Deep Learning algorithm that are loosely modeled after how
neurons in the human brain behave. They are a system for organizing machine learning algorithms
to perform specific tasks. A key advantage is their ability to automatically extract data features
without programmer input, making them fast and efficient for large datasets like images.
A neural network is structured into multiple layers, each with several nodes. The three main layers
and their functions are:
Input Layer: This is the first layer, whose primary job is to acquire data and feed it into the
network. No processing occurs at this layer.
Hidden Layers: These are where all the processing occurs. Each node in a hidden layer executes its
own machine learning algorithm on the data received from the input layer. Computations use weights
and biases, and information passes to the next layer after going through an activation function. The
network learns by adjusting these weights based on the error between desired and actual output. There
can be multiple hidden layers depending on complexity.
Output Layer: This layer receives the final processed data from the last hidden layer. Similar to the
input layer, it does not process the data. Its purpose is for the user-interface, providing the final
result to the user.
Q5. Differentiate between classification and regression model.
The difference between classification and regression models, both sub-categories of supervised learning
models, can be understood as follows:
Classification Model:
o In a classification model, data is classified according to
predefined labels.
o This model typically works on a discrete dataset, meaning the
output data need not be continuous.
o The problem usually involves predicting a specific type of
class label as the output from given input data.
o Examples:
Classifying students based on grades obtained from examination marks.
Predicting whether tomorrow's weather will be "hot" or "cold" based on historical
data.
Classifying emails as "spam" or "not spam".
Identifying if a customer is "eligible" or "not eligible" for a bank loan.
Regression Model:
o Regression models work on continuous data.
o These algorithms predict a continuous value based on
input variables.
o Examples:
Predicting your next salary based on previous
salary and increments.
Predicting temperature.
Predicting the price of a house based on parameters like the number of bedrooms,
carpet size, and garage area.
Predicting the selling price of a used car based on factors like fuel type, years of
service, and kilometers driven.
Predicting weather for the next 24 hours, as it will be a continuous range
Q6. What is neural network? Give the functioning of its three layers?
Neural Networks are Deep Learning algorithms modelled after the human brain's neurons.
They are systems organizing machine learning algorithms to perform tasks, notably automatically
extracting data features and solving problems efficiently with large datasets like images.
A neural network is structured into multiple layers, each with nodes. The three main layers and
their functions are:
Input Layer: The first layer, it acquires and feeds data into the network with no
processing.
Hidden Layers: These layers perform all the processing. Each node runs its own machine
learning algorithm, using weights and biases for computations. The network learns by
adjusting weights based on output error. There can be multiple hidden layers depending on
complexity.
Output Layer: Receives the final processed data from the last hidden layer. Like the input
layer, it does not process data; its purpose is for the user-interface to provide the final
output.
Q7: Convert the following scenarios to perceptron:
Context: A manager is deciding whether to approve a work-from-home request from an employee.
Factors:
-Does the employee perform well when working remotely?
- Are there any upcoming team meetings or collaborative projects?
- Does the company's policy support remote work?
- Is it beneficial for both the employee and the company?
X1
W1
y
X2 W2
Summation
&
W3 Activation
X3 function
X4 WB
W4
B
Performs well remotely
W1=0.5
Approve work-
from-home request
No upcoming meetings or W2=0.3
collaboration needed
Summation
Deny work-from-
&
W3=0.1 home request
Activation
Supports remote work function
Bais=0.7
Beneficial to both employee
and company W4=0.4
threshold
Performs well remotely
1 W1=0.5 Y: Approve work-
from-home request
No upcoming meetings or
1 collaboration needed W2=0.3 Deny work-from-
y=step(0.5(1)+0.3(1)+0.1 home request
(1)+0.4(1)−0.7)
Supports remote work W3=0.1 =step(1.3−0.7)
1 =step(0.6)
Beneficial to both employee W4=0.4 Bais=0.7
1 and company
threshold
From the calculation the output is 0.6. Since this is higher than the threshold (which is zero), the
result is manager can approve a work-from-home for an employee