Tuning hyperparameter with Akaike information Criterion

[melitong]

Hello, i am new in keras-tuner. I have tried hyperparameter tuning with keras-tuner to maximize the accuracy of validation set.
However, I have read about Akaike Information Criterion (AIC), it can be used to select best model. Can i use the AIC as a metrics or objective function in hyperparameter tuning and get best model with minumum AIC in keras-tuner?

[haifeng-jin]

Yes, you may need to use objective=kerastuner.Objective('val_aic', direction='max'), and define a custom metric named 'aic' and pass to the tuner.
We will add this tutorial to the docs.