5 ms·
I have tensors in later layers whose shapes depend on the training of the previous layers. Rad! Do you have any examples (or literature) that explains when thi
by vaughngh 9y ago
I have tensors in later layers whose shapes depend on the training of the previous layers.
Rad! Do you have any examples (or literature) that explains when this is beneficial?
- gormanc 9y agoNot yet! I'm not using convnets or backprop or anything so I don't think it would be beneficial that way, but you could get something similar to what I'm doing by looking at Fritzke's Growing Neural Gas[1] [1] http://papers.nips.cc/paper/893-a-growing-neural-gas-network-learns-topologies.pdf http://papers.nips.cc/paper/893-a-growing-neural-gas-network...
- vaughngh 9y agoNeat, thanks for the link.