ipazc / deepevolution Goto Github PK

At first congrats for the initiative, your implementation is brilliant!

I'm getting sporadically the same error in line:

weights_sorted = pd.Series(self._generation, index=scores).sort_index(ascending=False)

that is triggered from inside the series.py inside pandas, at the code:

        elif is_list_like(data):

            # a scalar numpy array is list-like but doesn't
            # have a proper length
            try:
                if len(index) != len(data):
                    raise ValueError(
                        f"Length of passed values is {len(data)}, "
                        f"index implies {len(index)}."

What I noticed is that this only occurs when index become (for some reason) a pandas Multindex instead of a list of arrays, it raises the following error message:

ValueError: Length of passed values is 14, index implies 1.

Thanks!

Hallo Iván,

first of all, thank you for your code, it is hard to find a good one for keras.
My first question is regarding the model validation. I did not quite find a way to implement a validaition dataset method in your code. Normally you generate a test/train dataset, a validation datasat and a testset. With your code, the model is trained on the train/test dataset AND validated against it. I would like to use a validation dataset for that use. In Keras you would implement this with validation_split or a pre-defined validation dataset. Is there a way to implement this without changing the deepevolution code itself?

Second question is about extracting the best fitted model. Somehow keras does not use the best fitted model for other functions (model.summary; model.predict...) but only uses the initial model. Is there a special kind of model call needed?