what is Actually Forward Propagation?
First step of Training neural network let us start understanding using dataset
let's understand through examples:
Data Set overview:
let take an example and as marked in diagram :
let us consider the first-row of dataset and build a neural network so it will easy to visualize
Each and independent feature First row of a dataset will passed to input-layer to a neural network act as an input
start as forward propagation by randomly intialising weight of all neurons this weight affected by edges connecting to neurons
weight are thought of more apropriately between two layers since edges connecting two layers
as shown in the diagram will show red-colored
let's discuss about first neuron of first hidden layer this neuron will receive input from a previous layer and combines with own set of weights and represented by red edges After applying some Activation function calculates output for hidden activation for this neuron .
hidden Activation will serve as input for next layer neurons similarly for all hidden activation layer for all neurons are calculated
when hidden activation are calculated this is said to have calculated hidden activation for first layer
hidden Activation layer will serve as input to these neurons as described below in diagram
once last hidden layer is calculated and combined by a finite set of weight between last hidden layer and output layer to produce output single row of observation
actual value will be not validated with predicted value as you notice we have calculated activation sequentially neuron that means or one after the other
one optimization we can make is to calculate in parallel since calculation is independent of each other that can be computed in parallel easily.
we can calculate the first hidden layer easily as shown in the diagram below:
how we can compute parallel easily first layer then calculate the second layer simultaneously then finally calculate the output as shown in diagram below
final summary and definition derived for Forwarding propagation:
series of calculation from the input layer to the output layer is called forward propagation.