To Keras community: What interpretations do you have for these curves?

[sondosaabed]

Hello all,

The accuracy of validation becomes steady between 40-60 while the training accuracy is getting higher. The loss is at first decreasing then starts increasing in validation. Does that mean the model is learning or that the model is over fitting the data?

x_valid.shape
(3000, 180, 180, 1)

x_train.shape
(14000, 180, 180, 1)

[fchollet]

That's overfitting, your model is plateauing after ~10 epochs. Either stop training at that point, or add regularization.

Closing this since Keras issues are for bug reports and feature requests, rather than for general ML support questions.

[sondosaabed]

@fchollet alright, thank you!

[sondosaabed]

3755