#neural #network #deep #learning


A Rust library to train and infer deep learning models

8 releases (4 breaking)

✓ Uses Rust 2018 edition

0.5.1 May 17, 2020
0.5.0 May 8, 2020
0.4.0 May 4, 2020
0.3.0 Apr 28, 2020
0.1.0 Mar 31, 2020

#23 in Machine learning

40 downloads per month

MIT license

1.5K SLoC

Ruby 1.5K SLoC // 0.1% comments Scheme 16 SLoC // 0.2% comments


crates.io Downloads from crates.io Rust doc.rs MIT licensed

Newron is a Rust library to train deep learning models.

Projet is currently under heavy development. No stable API yet !

Please see the docs or review examples to learn more about Newron features.



  • Dense (Linear fully connected layer)
  • ReLU
  • TanH
  • Sigmoid
  • Softmax


  • Mean Square Error (MSE)
  • Categorical Entropy


  • SGD


Feature Progress Comment
Improve metrics (confusion matrix struct and method to compute accuracy, recall..) ⬤◯◯◯◯
Implement Convolutional Layer ◯◯◯◯◯
Implement RNN Layer ◯◯◯◯◯
Webapp to monitor training phase ◯◯◯◯◯
Save and load trained models/weights ◯◯◯◯◯


Your contribution is highly appreciated. Do not hesitate to open an issue or a pull request. Note that any contribution submitted for inclusion in the project will be licensed according to the terms given in LICENSE.


This SOFTWARE PRODUCT is provided by THE PROVIDER "as is" and "with all faults." THE PROVIDER makes no representations or warranties of any kind concerning the safety, suitability, lack of viruses, inaccuracies, typographical errors, or other harmful components of this SOFTWARE PRODUCT. There are inherent dangers in the use of any software, and you are solely responsible for determining whether this SOFTWARE PRODUCT is compatible with your equipment and other software installed on your equipment. You are also solely responsible for the protection of your equipment and backup of your data, and THE PROVIDER will not be liable for any damages you may suffer in connection with using, modifying, or distributing this SOFTWARE PRODUCT.

No runtime deps