Making sense of blocks python package output


I've modified the Blocks tutorial, to train an MLP neural net with a dataset provided for an assignment in my ML course.

I'd like to evaluate the accuracy of the network with varying parameterization, but I am not sure how to obtain the accuracy in first place.

Inspecting the main_loop object useing dir() & vars(), I'm not coming across anything other than the test_cost_with_regularization.

It is possible to record means of different Theano variables via the monitor classes, so perhaps the answer lies there within?

Final Output

Training status:
     batch_interrupt_received: False
     epoch_interrupt_received: False
     epoch_started: False
     epochs_done: 1
     iterations_done: 16
     received_first_batch: True
     resumed_from: None
     training_started: True
Log records from the iteration 16:
     test_cost_with_regularization: 2.3032071347
     training_finish_requested: True
     training_finished: True


Posted 2015-11-03T16:24:33.707

Reputation: 111

No answers