ml-neural-network

Machine Learning Project @UniPi

Exam done

Suggestions from the professor

• K-fold is needed
• The model is in underfitting, it can be improved

TO USE

• pandas : può essere utile per il preprocessing dei dati (mean removal, decorrelation, covariance equalization)
• numpy : per operazioni tra matrici e vettori così da avere un codice più pulito

TO DO

• backpropagation
• regularization
• momentum
• k-fold cross validation
• grid search for hyperparams
• MEE, MSE and others (i.e accuracy, etc...)
• plots
• on-line, batch, mini-batch training