Basic idea about Neural Networks

Deep Learning is a subfield of Machine Learning, concerned with algorithms inspired by the human brain. Deep learning is not a newborn technology. The history of Deep Learning can be traced back to 1943 when Walter Pitts and Warren McCulloch created a computer model based on the neural networks of the human brain.

Neural Network Basic Structure

How the Neural Network Works?

•The raw data is first fed to the perceptrons of the Input Layer. 

•The data along with the bias traverses along the middle layers, also known as Hidden layers till it reaches the terminal layer and gives a random output.

•The Input Layer works only as the gateway to the neural network, it is not responsible for anything except for just Inputs.

•Hidden Layers on the other side have the biggest role to play. In the middle layers, we have something in common in every node, which is an activation function.

•Various Activation functions like ReLu, Sigmoid, Leaky ReLu, tanh are responsible for the activation of a specific node.

•Now you must be wondering that from the last few minutes we have been talking about Human brains, neurons, and neural networks; why haven’t I talked about dendrites yet (Dendrites are nothing but connectors for the neurons). Well, that’s where the backpropagation and weight comes into play.

•We use the dendrons to denote the weights in the Neural Architecture. 

•The output of the previous node along with the added weight and bias is fed to the next node.

•This keeps on happening until we reach the last node and finally get the output. This output might not match with the desired one.

•So we use a cost function to find the error and make the errors backpropagate to adjust the weight and the bias till we find the correct result.

•And that’s how a Deep Learning model is trained.

Thank you for reading this article. Hope to see you in the next one.

Have a Good Day !!