Supervised Learning


In this post, we will discussion about the most commonly used technique in Machine Learning – Supervised Learning. From the above image, most of you might have got the idea what Supervised Learning is. In supervised Learning, the computer is taught to identify the pattern in the data to be processed. In short, we are teaching the computer to identify the pattern. The formal definition of Supervised learning according to Wikipedia is

Supervised learning is the machine learning task of learning a function that maps an input to an output based on example input-output pairs.

Basically, we are training the computer with input and output. On course of time, the computer will start to find the pattern that defines output based on the input. This is the known as the learning function which can be used further to predict output for new inputs.

Lets look at some examples to understand the supervised learning.

House Price prediction

Lets predict the price of the house based on the area in square feet. Honestly, I dont have any idea about the price of the house. Lets assume the data is relevant.

The graph shows the relation between the area of the house and the area. Lets assume that you are interested in buying a 750 square feet house. From the image it is clear that the house costs around 220K. But say, you have changed your mind and want to buy a house of 1750 square feet. So to derive the price, you drew the blue line that represent the relation between the area and price. From the image it is clear the cost is around 500K i. This is similar to supervised learning.

The program is given the input as area and the labelled output as price for learning. The program derives a function (blue line) that can be used to predict the price on new area given as input. Here the blue line represensts the linear function. But, you can see that the yellow curve is more in relation with the points in the image. So its more accurate to predict the values using the yellow curve as reference. This is a quadratic function.

For your information, the price of the house is a continuous value, even though we round the price to the nearest 10s or 100s. So we consider this as a regression example.

Cancer Predection

Now lets predict whether the tumor is cancerous or not based on the size of tumor. The image shows the relation betwwen the size of the tumour and whether it is cancerous or not. 0 represent non-cancerous tumors and 1 represent cancerous tumors.

All the non-cancerous tumors are represented as blue dots and the cancerous tumors as red cross. Now lets assume that the green dot repesent a tumor. We have have to find whether it is a cancerous or non-cancerous tumor. Its clear from the image that it is  a non-cancerous tumor.

Here, the main purpose of the example is to classify whether the tumor is cancerous or not based on the size. This is an example of classification problem. Here 0 and 1 are the labels for classification problem. A classification prblem can have multiple number of lables. The input for the program is the size of the tumor and the labelled outputs are the whether it is cancerous or not. The program should find a optimal value for the size of the tumor which can be used as reference for the classification for new inputs.

Hope that you understood the supervised learning technique in Machine Learning. In the next post, we will discuss about Unsupervised learning. Feel free to cooment on the post.

Previous : Machine Learning – Introduction

Posted by Joyal Baby

Leave a Reply