1

Using AutoKeras is simple: 

clf = ImageClassifier(verbose=True)
clf.fit(x_train, y_train, time_limit=12 * 60 * 60)
clf.final_fit(x_train, y_train, x_test, y_test, retrain=True)
y = clf.evaluate(x_test, y_test)

But what's the function of final_fit? Isn't the model already trained in fit? If I don't use final_fit, what is the consequence? Does evaluate gives a different value?

Another question: how can I get some information about the searching-process: number of models found, epochs used for each model, ... and some other statistics? Is the console output the only way?

Xinyu
  • 63
  • 11

1 Answers1

1

Final fit uses validation data during training once it figures out the best structure, basically you train on a little more data after playing around with hyperparameter tuning to get a slightly improved model based on your validation data size.

So do we need it, maybe not, depends on your dataset size, will it hurt the model, most probably not, should you use it, depends on the circumstances.

anand_v.singh
  • 2,768
  • 1
  • 16
  • 35