Skip to content

Latest commit

 

History

History
19 lines (15 loc) · 978 Bytes

File metadata and controls

19 lines (15 loc) · 978 Bytes

Machine Learning

Nowadays, large amount of structured and unstructured data is available.
This repository is to understand how ML algorithms work.

Supervised Learning: Classification vs Regression

Classification: supervised learning task with discrete class labels

Goal: Predict class labels of new instances, based on past observations.

Binary classification vs Multiclass classification

Regression: Prediction of continuous outcome

Goal: Fit a line to it minimizing the distance between sample points and the fitted line

Reinforcement Learning

The system (aka agent) improves its performance based on interactions with an environment.

  • Trial-and Error approach

The agent receives feedback (reward) from the environment.

This reward is not the correct ground truth. It is a sample experience.

Extensive interaction with the environment allows agent to learn a series of actions that maximizes this reward.