Based on what I understood that conv2d_grad_wrt_inputs used to apply deconvolution graph to derive the input from the activation map so we can check the (attention) the most attractive features in he input that activated the neurons. But I can't see the idea behind the second conv2d_grad_wrt_weights.
Kindly if I understood wrong or if you explain more I'll grateful. Thanks -- --- 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.
