Report for MLP MultiClass Classifier

Hyperparameter Tuning

As we can see for some hyperparameters the loss and the accuracy is very noisy, these are those where the model is always bypassing the minima due to bad hyperparameters
sgd-sigmoid-0.05 is overfitting on the train data just like minibatch-sigmoid-0.05
The best hyperparameters which are good both for train as well as the val dataset is : minibatch-tanh-0.01