Validation data is not properly splitted

djanloo / cmepda

An LSTM + Encoder network for UHECR AirShowers

1 stars 2 forks source link

Validation data is not properly splitted #8

Closed djanloo closed 2 years ago

djanloo commented 2 years ago

to be honest is not used at all

djanloo commented 2 years ago

Fixed using 70-15-15 proportions for train-test-val. Even if I do not agree with the choice (pd4ml) of a fixed splitting between test and train, the validation population is taken to be the second half of the test population.