> For the complete documentation index, see [llms.txt](https://tony-ng-1.gitbook.io/what-is-ml/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://tony-ng-1.gitbook.io/what-is-ml/types-of-learning.md).

# Types of Learning

## Types of Learning

There are **3** broad types of learning.

## Supervised Learning

This is the case where we **have** an *output variable* (y) acting as a supervisor. For instance, in the handwriting recognition scenario

* Where we have a large data set of alphabets with labels ranging from 'a' to 'z', 
* The labels acting as the 'supervisor' for *correct* or *incorrect* labeling of characters, which
* Ultimately aids in designing the *best hypothesis* (g) to predict the output variable (label) of the next instance of data

![Asian mom as output variable (y)](https://66.media.tumblr.com/tumblr_lyg6n5PwFg1qanfb0o5_250.gifv)

## Unsupervised Learning

This is the where we **do not** have the luxury of an *output variable* (y) acting as a supervisor. Hence, in the same handwriting recognition scenario,

* Given a large data set of characters, the machine is likely to **derive unknown clusters** of alphabets looking alike, 
* For example, "1", "i", "I", and "l" being in the same cluster for i.e. their symmetry / intensity (will be discussed subsequently)

![Cluster of puppies](https://media1.giphy.com/media/xIvXX8CPVRJu/source.gif)

## Reinforcement Learning

Similar to supervised learning, a *feedback* is given here. However, there is no data acting as explicitly as a supervisor. Reinforcement learning merely rewards and penalizes a model for making/deviating from its progress towards the desired output. 

For example:

* We can teach self-driving cars to drive by giving them rewards
* For instance where a positive result is given, we can give them a positive reward (score), and
* where an undesirable result is obtained, we can give a negative reward (score) to that

![ur model after training rasengan with computer for 91238123203 times ](https://media1.tenor.com/images/5c108f99358cb6504c819dcbfb15f7bf/tenor.gif?itemid=8206035)