Scribd
Upload
21 views2 pages

Machine Learning Concepts and Design

The document outlines the components of a well-posed learning problem, which include the task, performance measure, and experience. It details the five steps involved in designing a learning system, discusses various perspectives and issues in machine learning, and categorizes types of learning into supervised, unsupervised, reinforcement, and evolutionary algorithms. Each type of learning is illustrated with examples to clarify their applications.

Uploaded by

Praneetha Siripurapu
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
21 views2 pages

Machine Learning Concepts and Design

The document outlines the components of a well-posed learning problem, which include the task, performance measure, and experience. It details the five steps involved in designing a learning system, discusses various perspectives and issues in machine learning, and categorizes types of learning into supervised, unsupervised, reinforcement, and evolutionary algorithms. Each type of learning is illustrated with examples to clarify their applications.

Uploaded by

Praneetha Siripurapu
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

■ Machine Learning Notes

1■■ Well-Posed Learning Problem


A learning problem is **well-posed** when it satisfies three components: - **Task (T):**
The job the system needs to perform. - **Performance Measure (P):** How success is
measured. - **Experience (E):** Data or past actions the system learns from. Formal
definition: A program learns from experience E with respect to some class of tasks T and
performance measure P, if its performance at tasks in T, as measured by P, improves with
experience E.

Task (T) Performance (P) Experience (E)


Play Checkers Percentage of games won Self-play against itself
Handwriting Recognition Symbols correctly classified Database of words
Speech Recognition Utterances correctly recognized Audio samples
Robot Driving Average distance without errors Sensor images & commands

2■■ Designing a Learning System


Designing a learning system involves 5 steps: 1. Choosing the training experience
(direct/indirect, with/without teacher) 2. Choosing the target function (board → move,
board → value) 3. Choosing representation (tables, rules, polynomials, neural networks) 4.
Choosing a function approximation algorithm (gradient descent, LMS update) 5. Final
design (performance system, critic, generalizer, experiment generator)

■ Flowchart: Steps in Designing a Learning System

Training Experience Target Function Representation

Algorithm

Final Design

3■■ Perspectives & Issues in Machine Learning


**Perspectives:** - AI perspective: Learning = improving intelligence - Statistics
perspective: Learning = estimating functions from data - Neuroscience perspective:
Learning inspired by the human brain **Issues:** - How many training examples are
needed? - Impact of noisy data on learning accuracy - Complexity of hypothesis
representation - Prior knowledge and generalization - Theoretical limits of learnability

4■■ Types of Learning


1. **Supervised Learning** – Learns with labeled examples (input → output). Example:
Handwriting recognition, spam email classification. 2. **Unsupervised Learning** – No
labeled data, discovers hidden structure (clustering, association). Example: Customer
segmentation, topic modeling. 3. **Reinforcement Learning** – Learns via rewards and
punishments through trial & error. Example: Game-playing, robot navigation. 4.
**Evolutionary Algorithms** – Nature-inspired, survival of the fittest approach. Example:
Optimization problems, genetic algorithms.

■ Illustration: Types of Learning

Supervised Learning Unsupervised Learning

Reinforcement Learning Evolutionary Algorithms

You might also like