Supervised, Unsupervised and Reinforcement learning

As we all know supervised, unsupervised and reinforcement learning is three types of machine learning. For solving a specific problem we apply these three types of techniques. According to the data, prediction and requirements, we apply and collect the result. Here we will discuss these three in simple language

Supervised learning

Supervised learning is a technique in which we train the machine using labelled data. In terms of the flow chart, we could say this is the process

labelled data (known output) –> [data cleaning and EDA->algorithm->model making] (training) –> Model evaluation –> labelled output

Example:

In this kind of data where we have to predict the known data or in the test data we apply a supervised learning technique. For this type of problem, we solve it by using some algorithm. According to the data, we apply the different algorithms for model making process to get the result. These algorithms are linear regression, Logistic regression, Decision tree, Random Forrest, Naive Bayes classifier, K-nearest neighbour, and SVM. These algorithms are based on mathematical and statistical models and calculations. Obviously, the different algorithm has different model and calculation based on the data. (For more)

Unsupervised learning

It is the technique where we train machines using unlabelled data and allow algorithms to act information without any guidance. In this case, the output would be unknown and the data will be unlabelled. In the simple flow chart

Unlabelled data (unknown output) –> Understand patterns and discovering output –>Output (unlabelled)

Example:

Audio clips, images, video clips etc.

These kinds of data are unlabelled data. In this data, we have to understand the patterns. Basically, we have to understand the statistical properties of the data and cluster them into classes. For discovering the outputs. The algorithms used are K-means, C-means etc. In this process, data points convert into clusters and grouped the similar items clusters. (For more)

Reinforcement learning

It is a part of ML where an RL agent learns from the consequences of its actions by interacting with a space or an environment for observing rewards. In a simple flow chart, we could see a fixed action which is not being taught explicitly.

agent—(reward)(state)environment—agent—(reward)(state)environment…….

The agent is the RL algorithm, which learns from trial and error. The environment is the world through the agent’s moves. In this learning, some concepts are also applied like Reward maximization, in this theory we get to know about maximizing the reward by the way of training the RL agent for taking the best action. Exploration is about exploring and capturing information and Exploitation is exploited information which is known. The algorithm which we use is Q-learning, SARSA etc. (For more)

In the case of technology, we need to learn about the tools and libraries of the programming language. We need to learn about the algorithms of supervised, unsupervised and reinforcement learning and have to apply them according to the data and output.