subpath / neuro-evolution

Wrapper for neural evolution approach for hyperparameter tuning
MIT License
14 stars 6 forks source link

use neuro evolution for forecasting time series with LSTM #3

Open AliAzG opened 4 years ago

AliAzG commented 4 years ago

Hello Alexander, absolutely great work! my question is how can I use neuro-evolution for forecasting time series with LSTM? I wrote my code similar to your example but it gives me "best accuracy: 0.0"! do you know how can I fix it?

subpath commented 4 years ago

Hi! Thank you!

Unfortunately, the structure of the NN is hardcoded to be a feed-forward network, see train.py file. To run LSTM or any other types of network you need to change compile_model function