best optimiser and loss function and metric combination in deep learning code example

Example: loss funfction suited for softmax

model.compile(loss='categorical_crossentropy', optimizer=opt, metrics=['accuracy'])