Supervised, Unsupervised and Semi-Supervised Learning.
In this post, I will discuss what is Supervised, Unsupervised and Semi- Supervised Learning techniques and why to use it. Apart form that you will discover how this machine learning technique works and types of each technique.
First of all let’s talk about learning. What is learning ?
Learning is making use of past experience to improve the Future performance or to predict the future outcomes. Learning is making useful changes in the system that enables a system to do the same task more efficiently next time. For machine, experience comes in the form of data.
Machine learning is all about extracting important/relevant information from the data, learning form the data and applying that knowledge to analyze the new data. For any project we should always ask what is data and its dimensions? and what is the objective of the this project? Answer of question will help us to decide which machine learning we should as there are three technique.
1) Supervised machine Learning:
Generally speaking, most of machine learning uses Supervised learning.
In Supervised machine learning you are given data with correct labels and are asked to label new examples. Also, it usually requires a lot of data before it learns. Here the given data is called features (Input Data) which will be in the vector form in most cases and label(Predicted value) which will be in scalar form. In supervised learning you we train the machine with the data that already have a correct answers tagged(label). Machine learning model learn form the tagged labels and helps you to predict the output(label) of the unseen data. Due to this, it is important to provide unbiased data to the model. This will help to optimize the performance based on the previous examples.
Data points D = {d1, d2, d3,…, dn} is a set of n examples, For every data points di = <Xi , Yi > were Xi = input data vector and Yi = desired output. Here the objective is to approximate the mapping function (f: X → Y) comparatively so well that when we give new data (X) we can predict (Y) from previous experience.
Supervised learning is further divided into two types Regression and Classification.
Regression: In regression X (input) will be discrete or continuous and Y(output) will be continuous and real values.
Classification: In classification X (input) will be discrete or continuous and Y(output) will be discrete and categorical values.
Algorithms under supervised machine learning are Support vector machine, Neural network, Linear and logistics regression, random forest, and Classification trees.
2) Unsupervised machine Learning:
In Unsupervised machine learning, data is given without correct labels tagged and are asked to learn the correlation between the samples and components of samples in order to learn more about the data. Unsupervised methods help you to find features which can be useful for categorization.
Data points D = {d1, d2, d3,…, dn} is a set of n examples, For every data points di = <Xi> were Xi = input data vector. Here the objective for unsupervised learning is to model the distribution or underlying structure in the data.
Unsupervised learning is further divided into two types Clustering and Association.
Clustering: In clustering, similar types of data is cluster in groups.
Association: In association, rules are discover to describe the data.
Algorithms under unsupervised machine learning are Cluster algorithms, K-means, Hierarchical clustering.
3) Semi-supervised machine Learning:
Every time data doesn’t have the label tagged with them, there’re millions of data set in which some data points contains the label and other data points doesn’t have labels. Because of this semi-supervised learning can be positioned halfway between unsupervised and supervised learning models. The objective of semi-supervised model is to classify some of the unlabeled data using the labeled information set.
I hope you understand the difference between Supervised, Unsupervised and Semi- Supervised Learning. Good bye!
