Python: What Does train() Method in Pybrain Package Return?

Question

The link here says that trainer.train() returns

a double proportional to the error

What does that mean? I am using BackpropTrainer to train a neural network for classification. So far, my code has returned values less than 1. Does it mean that it validates the training data and returns ratio of number of misclassified samples to the total samples?

Answer 1

Reading the documentation it seems that it trains on the data for one epoch. It may be that it doesn't converge in that time, so it returns the mismatch (error).

trainer.trainUntilConvergence() will train on multiple epochs until it converges. It returns a tuple of each convergence mismatch.