use of validation data and model extraction about deepevolution HOT 2 CLOSED

Hi TobiasMerlin,

the implementation does not have the validation option available and it requires to change the code. Even though, you can use the evolve() generator and have a manual evaluation of the model between epochs, something like:

for _, best_score, mean_score, std_score in model.evolve(X, Y, max_generations, fitness_func, population, top_k, mutation_rate, mutation_std):
    validation = model.evaluate(X_val, Y_val)

Regarding the question of extracting the best fitted model, it is supposed that the best model is stored in the weights of the model at the end of every epoch. Maybe we can have a test example to check it out?

from deepevolution.

Hi Iván,

thanks for your answer. The implementation works:) And regarding the Problem with the Model extraction, it seems that was my fault. I made a mistake withe the Model callbacks. So i will close this comment

from deepevolution.