Skip to content

Conversation

@FlixCoder
Copy link

At first, I did not plan to create a pull request and to keep everything clean and abstract. But I think the improvements got really useful, so I decided to clean up and update the doc and README accordingly, but I don't know, if everything is still good enough. If there is something minor not ok, please tell me, so I can fix and you can merge safely.

@FlixCoder
Copy link
Author

Oh sorry, I forgot to add the .trayis.yml. It is in now.

@FlixCoder
Copy link
Author

This fork adds L2 regularization and several activation functions to the original crate. Additionally, there are a few minor improvements. Lambda can be set just like the learning rate. The activation functions for hidden and output gets set in NN::new as second and third parameter respectively.

@brickman1444
Copy link

@jackm321 I'd like to see this PR merged. I tried using the nn crate in a WebAssembly project but it failed to build because nn relies on the outdated rustc-serialize package which doesn't compile for WebAssembly (See wasm-bindgen/wasm-bindgen#1488). This change would fix that problem.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants