Introduction

Introduction

Definition of Learning: The ability to improve behaviour or a prediction based on experience.

Building comp sys that improve with experience

Machine Learning

• Explores algorithms
  • learn from data and build models from data
  • Models can be used for some tasks, eg prediction, decision making or solving
  Formal definition of ML (Mitchell): A computer program (machine) is said to learn from experience E, with respect to some class of tasks T, and performance measure P, if its performance on task T as measured by P improves with experience E.

Components of a Learning Algo

• Tasks \( T_i \)
  • Prediction
  • Classification
  • Acting in an environment
  • et cetera
• Experience E, also called Data
• Measure of improvement or P

Components of an ML Model

A ML model consists of a learner and a reasoner

• Learner
  • Takes up Experience Data and backgreound knowledge, and creates one or more predictive model(s) and provides it to the reasoner
• Reasoner
  • Takes up the task, applies the model it was provided to by the Learner to give a result/answer/performance

Learner

1. Choose the traning Experience or Data
2. Choose the target function [how to represent the model] that needs to be learned
3. Choose the class of Function [How to reperesent the target func] or the hypothesis language
4. Choose a learning algorithm to infer the target function

[Note to 2: Richer a representation is, tougher to learn the model]