Machine Learning Introduction

Course on Artificial Intelligence

Somnath Hazra

What is Machine Learning?

Tom Mitchell [1997]- A computer program is said to learn from experience E with respect to some class of tasks T and some performance measure P, if its performance at tasks in T, as measured by P, improves with experience E.

• For example, a learning algorithm on seeing enough number of labelled images of a dog is able to predict if the next image is of a dog with some accuracy.
• E is the act of training with many images of dog
• T is the task of predicting whether the unknown image is of a a dog
• P is accuracy

Machine Learning

Mainly 3 types of tasks:

• Supervised Learning- Learn from labelled data
• Unsupervised Learning- Tasks for unlabeled data
• Reinforcement Learning- Not explicitly labelled data but guided by incentives

What do machines learn?

Why do we need it?

• Systems to automatically adapt and customize according to users
• Personalised recommendations or advertisements
• Discover new knowledge from large databases
• Data mining
• Ability to replace human in monotonous tasks requiring some intelligence
• Develop systems that are too difficult to define and construct manually
• Image Classification system
• Navigation

Sample tasks

• Recognizing patterns
• Facial expressions
• Handwritten or spoken words
• Medical images
• Generating patterns
• Generating images
• Recognizing anomalies
• Credit card transactions
• Sensor readings at power plants
• Prediction
• Stock market

The No Free Lunch Theorem

Generalization capacity of algorithms from a finite set of training data

The No Free Lunch Theorem states:

Averaged over all possible data generating distributions, every classification algorithms has the same error rate when classifying previously unobserved points

​

In the absence of domain knowledge, no learning algorithm is universally better than any other

​

Therefore important to understand the distributions relevant to the domain

Thank you