FAQ

• How to mitigate the overfittng?

  • Try dropout of True , larger l2_regularization, less the number of node.

• How to restart from previous training?

  • Write the file name of checkpoint of potential_saved in neural_network of input.yaml. Do not forget to copy scale_factor and pca when using potential_saved.