Validation accuracy remains 0. about keras-extra HOT 3 CLOSED

Had a small problem with the loader, but even since it has been fixed I have been unable to effectively train a Neural Network!

from keras-extra.

Some might say that the "dark art" of deep learning is that you have to try a lot of things before something starts to work. It could be an issue with whether you are randomizing your validation data, or your initialization, or your optimizer. I notice for the optimizer at least you are using rmsprop but not the default parameters for it. Default rmsprop works for fine me, as well as default adam, when I train a CNN-LSTM. But, of course, I'm not very familiar with your data, so it'll really come down to experimentation and tweaking, but those are important things to look for when you're starting out. (I do think most likely it has to do with a bug in how you're loading and processing your data than anything else).

from keras-extra.

Quick update, you were correct. Some of my labels were incorrect, and as a result the NN was unable to distinguish between either class.

from keras-extra.