Hi guys, so now we are dealing with Perceptron. So we have just started deep learning and to be honest Perceptron is actually the building block of deep learning. So in future, whatever neural networks you read, one single unit is always the Perceptron.
And that is why reading Perceptron properly is very very important. So now... Perceptron's basic idea is that you already know that the Perceptron is a model based on neuron which is an individual neuron in our brain, based on that neuron So this is a biological neuron, so it has got this body, these dendrites and this exon So in a way, input comes from here and goes from here same thing to formulate we have created this model so in a perceptron what happens is in a perceptron what happens is that you have different inputs these are the inputs these are your actual inputs this is known as bias we will talk about this later this is known as bias and all of this is fed into the main body part of the person one more thing, this is called connection and all these corrections are weighted means, the impact of x1 will be decided by w1 the impact of x2 will be decided by w2 and so on now eventually what happens is you do a summation like this so what you do is, you do a summation like this you do this you find out the sum which is actually equal to w0 multiplied by 1 which is equal to w1 is this way plus w1 x1 plus w2 x2 plus wn xn. Okay.
Let's call this value to be equal to x. Okay. Now what you do is, see at the end what you want is result 1 or 0. Right.
You want to do classification at the end. So you take all these inputs and eventually you get result 1 or 0. But here you don't want to put this number 1 or 0 in between. So you actually need some sort of function which will pick up any number in any range and will enter 1 and 0. So what we do is we take an activation function. We call this activation function.
Activation function. And here also we take an activation function called step function. And step function works like this.
Here f of x would be 1 if x greater than 0 and 0 if x less than 0. So this means if this whole number is greater than 0, then you will say output is 1. If this whole number is less than 0, then you will say output is 0. And that is how perceptron works. So there are two things, first you do summation and then you send this result to this activation function You can change the activation function, there are other activation functions There is sine activation function, there is sigmoid Then there is ReLU, there is TANNage There are many functions but generally speaking if you are talking about Perceptron We use either the step function or the sign function Now what we will do is I planned to teach this playlist in this way So we saw the introduction in this part, very basic introduction Next we will see mathematical formulation After that we will make code of perceptron with scratch, in python Then we will see the implementation of SQL, how Perceptron is implemented in Psychic Learn And finally we will see what was the problem in Perceptron Because of which Perceptron could not become that important And why then the need for Multilayer Perceptrons increased, MLPs Which are basically your Artificial Neural Networks So I guess you know that there is a problem in Individual Perceptrons Because of which MLPs have to be invented Okay we will discuss everything in the upcoming videos, I know till now the discussion is clear, thanks for watching.