Created a four layer Neural Network that works on the MNIST Fashion dataset. This code builds a 4-layer neural network with 256 hidden nodes per layer except the last layer which has 10 (number of classes) layers. I use Minibatch Gradient Descent, which runs for a given number of epochs and does the following per epoch: Shuffle the training data Split the data into batches (use batch size of 200) For each batch (subset of data): feed batch into the 4-layer neural network Compute loss and update weights Observe the total loss and go to next iteration
The code can be run by running python nn_main.py