I'm working on the network of the deeplearning LeNet and I applied the reul
to the convolution. I'm not sure if all the parameters are included in the
gradient will consider every function in the graph.
the convolution is:
conv_out = conv.conv2d(
input=input,
filters=self.W,
filter_shape=filter_shape,
image_shape=image_shape
)
rectified_conv_out = T.nnet.relu(conv_out + self.b.dimshuffle('x', 0, 'x', 'x'))
pooled_out = downsample.max_pool_2d(
input= rectified_conv_out,
ds=poolsize,
ignore_border=True
)
while the gradient is
updates = [
(param_i, param_i - learning_rate * grad_i)
for param_i, grad_i in zip(classifier.params, grads)
]
and I did save all the parameters in params.
--
---
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.
