In the last video we have seen how a neural network works in a man. Now we are going to see how to create an artificial neural network.
Already you know that a set of neurons connected together to form a neural network to make a decision for doing a job.
Here we are going to have instead of neurons we are going to have set of nodes. This nodes will receive multiple input and after that it will go the next nodes that is called a layer and another layer and n number of layers and finally it gives the output . This is called an artificial neural network.
Now I will teach you how to build a neural network. Say for example if you have a football, if a ball is shown to a person then he will be able to identify a football.
May be it looks for a human being for a machine you have to tell the machine to learn that it is a football. That is a challenge here.
Now What I am going to this neural network the input will be the shape of the ball, the size of the ball and what is the colour of the ball and what is the diameter of the ball. Now we will have the input layer there we are going to give the four inputs that is colour,shape, size, weight and the size of the ball.
The colour will be identified in the next layer, if the colour is black and white the weight number is assigned as 1 and all other nodes will not assigned any number. This layer only identify the colour of the ball
The next layer it will identify the shape of it and the next layer it will identify the weight of the ball.
For a foot ball the colour assign a number , for a weight I ll assign a number and the shape I’ll assign a number are assigned a number and all the the numbers are added and the sum of the number should be equal to is 1 if you assign 0.25 for each input, if it is equal to 1 then this program will identify or this neural network will identify that this is a football. clear
So For a football we are assigning a number and that number is added together and the final value is given to the computer if it is 1 then it is identify it is a football. Clear
Now let us see I am going to teach this neural network how to identify a cricket ball. What is the size of the ball will be given what is the weight of the ball what is the colour of the ball and the what is the shape of the ball. For a cricket ball I will give a different number, if I give 0.25 for football , I will give 0.1 for cricket ball. If all the measure is given as 0.1 then the therschold value will be 0.1+0.1+0.1+0.1+0.1 then it will be 0.4 .
If the output is 0.4 then it is a cricket ball. Now this neural network is able to identify a football and a cricket ball.
There will be a problem if I introduce a tennis ball. If it is a football then the number is 1 and if it is a cricket ball 0.4 .
Now I am giving a tennis ball , the size and shape of the tennis ball will be equal but the weight and colour will be different. So What happens is if you show the tennis ball then it will say it is a cricket ball. So this machine does not learn the tennis ball. Now we have to teach it is a tennis ball now. And it has The shape and size are different for that assign a different number so it will get a different number . so like this you cab to Teach for all the balls then the computer will be able to tell the name of the balls correctly then we are successful.
But it is difficult. Do not think that you will have to a large amount of data feed the data into the network then you will be able to get a correct output. That is not possible You have start with one or two datalike this and keep on teaching the computer . This is called supervised learning .and the computer have supervised learning for computer so that the computer will be able to identify three balls correctly . Keep on increasing then it will be able to get correct answer.
Is it clear?
Now This is for only to see for example , we have seen an example of automatically self driving car. Now we will be able to understand the attributes this machine to and what are the identification a self driving car should have .okay
It is has to identify all the human being, all thecars all the vehicle , cycle and the signal.
So all the input has to go inside , for each one a neural network will be there and all these combined together it will form a big neural network, working successfully then , if that car is able to drive a car a numerious networks work put together and it works successful . It is complex and writing a program also complex so you have to start with smallest thing as identifying object and combined is so many things then you will be able to build a neural network . When you are able to build a neural network and it is able to do it successful combine a each neural network into a bigger neural network .
As we are having self driving car, identifying a man is a small programm , identifying another car is a programme and identifying a signal is another small programmes, you combine all the small programs or neural network it will do the magic that it will drive the car by itself.
