CNN + Required DualArray implementation #11
+472
−76
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
sorry for the big pull request, I just realised that for some reason Github no longer recognised my copy of DualArrays.jl as a fork so I created a new fork and moved everything over.
This is a working CNN implementation and the DualArray/DualMatrix implementations required. I have commented motivations for things like multiplication. I have a couple of questions:
-Right now it seems a bit slow: although after a few hundred epochs the network seems to guess the MNIST numbers pretty consistently, it takes 1-2 seconds per epoch (I can implement this in Lux and compare speeds). I suspect this is because sparsity of the Jacobian is not preserved through converting to and from 4-tensors. What tools are available to help with this?
-I have changed partials of the Dual type to be an array instead of a vector to preserve directions of perturbations when indexing from a DualMatrix. Is this the right thing to do?