I am currently experimenting with the best combination of: number of layers, nodes per layer and number of epochs for training (using adam optimizer, dense layers and a combination of relu and softmax activation functions)
In the feature I will be experimenting with different optimizers, layer types and activation funcitons