Convolutional RBM

In his video about collaborative filtering, Hinton talks about a big RBM network that is actually many RBMs that share weight. (In the netflix competition, if two users rated the same movie then the weights from the visible units (movie) to the hidden units are the same).

Another network we saw in class that shares weights is a convolutional network. What if we apply this idea to RBM. Instead of having fully connected hidden units, each hidden unit is connected to a subset of the visible unit. Like in convolution, other hidden units are connected to a different subset of the visible unit but they share the same weights.

Has this been tried before ? Like for convolutional network, does it give better results for images ?


0 Responses to “Convolutional RBM”

  1. Leave a Comment

Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: