Types of ML Models: Supervised, Unsupervised & Reinforcement Learning.

There are several machine learning methods to train models for accurate outcomes. Prediction models are evolving with new machine learning methods that are complex algorithms that need to explore and understand by even expert data scientists.

These machine learning models are classified broadly into:

1. Supervised Machine Learning

Supervised learning techniques are those models that do predictive analysis among input and output variables. The outcomes are definite since input variables are known. There can be multiple input variables to map functions for an output variable

Where you have an input variable (x), output variable (y), use an algorithm to learn the mapping function from the input to output Y = f(X).

Supervised machine learning includes two major processes: Regression and Classification.

Regression:

Regression is the kind of predictive analysis that investigates the relationship between the response (dependent) variable and explanatory (independent) variable. The output variable(y) is a real or continuous value, such as price and weight variables. Here we will find the correlation between the input and output variables.

Classification:

A classification problem is when the output variable is a category, such as “Male” or “Female” and “Defaulters” or “Non-Defaulters” variables. A classification model attempts to draw some conclusions from observed values. Given one or more inputs, a classification model will try to predict the result of one or more outcomes. 

Widely used Supervised learning algorithms are:

2. Unsupervised Learning

Unsupervised learning is the training of the model using information that is neither classified nor labeled. Unsupervised learning is self-learning techniques where the algorithm can find previously unknown patterns in a dataset that does not have any labels.

Classes are highly similar to a single class type and a bit similar to other classes.

Unsupervised learning classified into two categories of algorithms:

Clustering: A clustering problem is where you want to discover the inheriting grouping in the data, such as grouping in the data.

Association: An association rule learning problem is where you want to discover rules that describe a large portion of your data, such as people that buy X also tend to Y.

Widely used Unsupervised learning algorithms are:

3. Reinforcement Learning

It is learning by interacting with space or an environment. An RL action is based on its experience and also by new choices.

Reinforcement is a type of machine learning, where we use it to make a sequence of decisions. It also helps us to discover which action yields the highest reward over a long period.

Consider a scenario of teaching new tricks to your pet. As your pet doesn’t understand human language, we can not direct the pet directly what to do. We emulate a situation that the pet tries to respond in many different ways. If the pet’s response is the desired way, we will give a reward to it (food). Now whenever the pet encounters the same situation, the pet executes a similar action even more enthusiastically in expectation of getting more reward. That is like learning that pets get from “What to do” from positive experiences. At the same time, the pet understands what not to do when faced with negative experiences.

In this case, your pet is an agent that learns the environment. In this case, it is your house. An example of a state could be your pet sitting, and you use a specific word in for cat to walk. The trained model responds by executing an action change from one state to another. For example, your cat acts lying to walking. The reaction of an agent is an action, and the policy is a method of selecting an act given a state in expectation of better outcomes. Change in action will help to get a reward or penalty in return.  

There are three approaches to implement the reinforcement learning model: 

1. Value-Based 
2. Policy-Based 
3. Model-Based

Differences between supervised learning and unsupervised learning:

Supervised Learning	Unsupervised Learning
The supervised learning goal is to determine the function so well that for a new input dataset, it will predict the output.	The unsupervised learning goal model the hidden data patterns or underlying structure in the given input data.
Explanatory variables, target variables are provided.	Only an Explanatory variable is input to the system.
Supervised learning methods utilized in applications like image and speech recognition, forecasting, financial analysis, etc.	Unsupervised learning algorithms help to discover interesting relations between variables in large databases.
Examples: Classification and Regression	Examples: Clustering, Anomaly detection, Etc.