Hi all. I'm trying SdA implementation from http://deeplearning.net/tutorial/SdA.html Execution of the code test takes 1146 minutes with my MacBook Pro 7,1 (Intel Core 2 duo), more longer than time reported by the tutorial.
Suppose I use a grid search of 1000 elements equally distributed on parameters. So, time for completion is very high. An idea is use only model which test parameters during model selection, but its not a good idea, because we know that its very aleatory and its strictly depends only on initialization of weights and bias. Is there a fashion, a way to reduce time completion of the pretraining and finetuning? Or is there a way to improve model selection? Thank you -- --- You received this message because you are subscribed to the Google Groups "theano-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
