I am using MLPClassifier. The model consists of 3 hidden layers each of 128 neurons. The input dataset has 1 million points each of dim = 128. After playing around MLPClassifier parameters, I still get poor prediction accuracy. I am not sure if this is caused by underfitting or not! Can underfitting be caused by the lack of enough data during the training process?

Thank you


Did you compare train/validation loss? If both are large it could be underfitting. Although, 1 million points seem like a lot of data.

To make sure that your code is correct I would suggest to overfit the model - pick a small part of your dataset and see if train loss goes to zero (and accuracy goes to 1).

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.