NeuPy is a Tensorflow based python library for prototyping and building neural networks
Main changes:
repeat
functionMain changes:
define
method for all layersMain changes:
>
(still supported) to >>
and lists to|
operator for parallel connectionsexponential_decay
and polynomial_decay
signals
attribute)GroupNorm
layerpredict
methodsshow
method to the networksSmall changes:
error
parameters for optimizers was renamed to loss
training_errors
and validation_errors
and it were replaced with new errors
attributeprediction_error
method to score
Misc:
Removed
train_end_signal
and epoch_end_signal
plots.error_plot
functionplots.network_structure
functionChanges:
Changes:
Changes:
compile
method was removedRemoved:
Bugs:
Features:
Bug:
epoch_time
in order to prevent crashes when training was stopped on first iteration.Fixes: