In this video, I am going to tell you what machine learning is. Machine learning is a tool for AI. It is nothing but teaching the computer to learn and relearn and it has to improve on learning. That is what machine learning is. Let me explain to you how it works. There are three things involved in this+. One is input A, output B and the mapping between A and B. We are going to teach the computer input A, output B and the pattern between or the mapping between A and B will be taught to the computer so that it will be able to identify whenever you are giving A, B will be given as output.
I am going to teach the process of identifying dog or a cat. If a picture is given as input, this process has to tell us whether it is a dog picture or cat picture. What are we going to do?
We are going to give a training set of five pictures of dog and five pictures of cat to this program. And we are also going to tell how to get the pattern of the eyes and pattern of the mouth and pattern of the ears. You know that the pattern of the mouth, pattern of ears and the pattern of eyes for dog and cat are entirely different. So we will teach them if this pattern is there, then it is cat and if this pattern is there, then it is dog. So this computer will be able to identify when any picture is given based on the pattern it will be able to identify whether it is a dog or cat.
Now you might be thinking what the difference between ordinary programming and this machine learning program is. In the ordinary program, if you are giving the training set, say for example, five pictures of dog and five pictures of cat, then it will be able to identify only when those five pictures of dog and five pictures of cat. It will not be able to identify any other pictures of dog and cat. But in this machine learning process, it is able to identify any picture of dog or any picture of a cat because this we are teaching them to understand the pattern of the eyes, pattern of the mouth and pattern of the] ears of dog and cat. That is the difference between ordinary programming language and AI program.
This machine learning, if you have five set of data, this is small number of data. Now if I am giving picture, say if I give 10 pictures, it may be able to identify seven or eight pictures correctly, three or four pictures it may get it wrong because that pattern will not be available in this five pictures. Now if you want to improvise, or improve the accuracy of detecting whether it is a dog or cat picture, then we have to increase the number of pictures, the input has to be increased.
For this, we need big data. What is big data? Big data is nothing but large volume of data, large variety of data, which is available in the internet and we can store this data because of the technology development taken place. So we can store data. In our example, we have dog picture and cat picture which are available in the internet in plenty. So we can store this data. The patterns can be learned by the computer. So the accuracy rate will be increasing. Because of this large data, big data, the accuracy rate – we can raise up to 90% to 95%. Now I want to get it to 100%. How is it possible? It is possible by using another tool – AI tool – which is called deep learning. Deep learning can be done through neural network. We will learn about neural network and deep learning in the next class.
