Compared to forward convolution conv2d_grad_wrt_inputs is very slow if it is implemented with several deep layers compared to forward convolution. In case you are implementing attention finding in each image it will take long process time and it is convenient.
-- --- 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.
