Deep Learning CNN for clothing classification using Fashion MNIST dataset Neural Network was implemented using Python and PyTorch.
The NN was implemented following the architecture below.
The model uses a cross entropy loss function and SGD as optimisation method. Weights are initialized using the Xavier normal initialization.
Trained the model for 50 epochs with the following activation functions: Sigmoid, reLU and Tanh.
Results:
Activation Function: Tanh
Learning rate: 0.1
Activation Function: reLU
Activation Function: Sigmoid
