Learning Styles

Supervised

The algorithm will use training dataset containing input data (your predictors) and the output value you want to predict (which can be discrete or numeric) to learn a link between input and output.

Underlying idea is to identify a ++generalized or approximated++ function that maps input to output. With this function, a prediction model will be formed to predict output for new data with some accuracy.

When to use: To analyse a fully labelled (the value to be predicated for new input) data set

Unsupervised

On the other hand, unsupervised learning does not use training data. It learns from raw input data which is not labeled or without an attribute that we want to predict

There is no well defined prediction output. The goal is to organize data by similarity or finding a hidden pattern in the data by a mathematical process by ++deducing structures++ present in the input data

Another example

When to use: To pre-process the data, during the exploratory analysis or to pre-train supervised learning algorithms.

Semi-Supervised

A combination of both Supervised/Unsupervised approach. Input data is a mixture of labeled and unlabelled examples, hence must learn the structures to organize the data as well as make predictions.

One of part of algorithm, called the generator, tries to create new data points that mimic the training data. The other part, the discriminator, pulls in these newly generated data and evaluates whether they are part of the training data or fakes. The networks improve in a positive feedback loop — as the discriminator gets better at separating the fakes from the originals, the generator improves its ability to create convincing fakes.

When to use: This method is particularly useful when extracting relevant features from the data is difficult, and labeling examples is a time-intensive task for experts.

Reinforced

Algorithm (aka Agent in this case) focus on the end goal over structure of data. We use cues to train a dog to perform certain tasks and reward the dog for its action

Reinforcement learning operates on the same principle, AI agents (Mouse) are attempting to find the optimal way to accomplish a particular goal (finding nearest route to cheese), or improve performance on a specific task. As the agent takes action that goes toward the goal, it receives a reward. The overall aim: predict the best next step to take to earn the biggest final reward.

When to use: ??

References:

1. Nvidia Blog
2. machinelearningmastery