Table of Contents  ▸  List of Commands  ▸  Neural Networks  ▸  nn_trainer

◀  nn_transpose    |    nn_varname  ▶

nn_trainer

Arguments:

• out,loss,_optimizer,_"optimizer_params",_scheduler,_"scheduler_params"

Description:

Add a network trainer to the current network, with specified optimizer and scheduler.

optimizer can be { adam | adamax | dtschump | sgd | rmsprop }.
scheduler can be { constant | linear | exponential | adaptive }.
adam, adamax: (learning_rate,beta1,beta2,weight_decay,clip_gradient_min,clip_gradient_max).
dtschump: ( learning_rate,momentum,weight_decay,clip_gradient_min,clip_gradient_max ).
rmsprop: ( learning_rate,momentum,beta,weight_decay,clip_gradient_min,clip_gradient_max ).
sgd: ( learning_rate,momentum,weight_decay,clip_gradient_min,clip_gradient_max ).
constant : (none).
adaptive : ( lr_min,lr_max,nb_warmup_iterations,nb_cooldown_iterations,patience_down,factor_down,patience_up,factor_up ).

Default values:

optimizer=adam, optimizer_params=(default), scheduler=constant and scheduler_params=(default).

G'MIC is an open-source software distributed under the CeCILL free software licenses (LGPL-like and/or
GPL-compatible). Copyrights (C) Since July 2008, David Tschumperlé - GREYC UMR CNRS 6072, Image Team.