How do we get a valid output in Neural Network?

 What we are gonna do is, we are gonna train the network with a ton of different information. So,let’s say we play 1000 games of snake and we get all of the different inputs and all of the different outputs. We consider did the snake survive or not.

What we will do is, we will train the network and what it will do is, it starts adjusting it’s biases and it’s weights to properly get a correct output.

    If the network got the correct output the network is not gonna bother adjusting the weights and network. But let’s say we go the output wrong, now what the networks will do is it will start adjusting these weights and adjusting it biases by adding or multiplying by a constant value.

Ex: b1 +1, w1 – 1 ( b – bias, w - weight)

        So, that’s why typically take a massive amount of information to train. So,what we do is pass all of this information and it keeps going the networks. And in the beginning it sucks because the networks sets a random weights and random biases. 

But as it go through learns. It is gonna leave the network as it is if it got the correct output, else it is gonna adjust the network. Eventually your goal is where you get a point where your network is very accurate because it is given a ton of data and it adjusted it’s weighted and biases correctly. So that the output will be accurate.