Neural networks are mathematical expression which take inputs as data and the parameters of the neural network which are weights, bias, the mathematical expression for the forward pass (the process of feeding input data through the network’s layers to produce an output).
This is followed by loss function which measure the accuracy of the prediction, it will be low when prediction are matching the target.
Then how to decrease the loss function, by using backptopegation we get the gradient and these gradients tell how to tune the parameters to decrease the loss.
Biases gives neural networks the flexibility needed to model a far richer class of functions than if they were forced to always pass through the origin.
A simple example of this is a perceptron