# Another Neural Network Trainer

**URL:** <https://discourse.processing.org/t/another-neural-network-trainer/22624>\
**Category:** Gallery\
**Created:** [July 14, 2020, 9:53pm UTC](https://discourse.processing.org/t/another-neural-network-trainer/22624 "2020-07-14T21:53:09Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![ivaylo](https://avatars.discourse-cdn.com/v4/letter/i/5fc32e/32.png) [@ivaylo](https://discourse.processing.org/u/ivaylo)\
**Post date:** [July 14, 2020, 9:53pm UTC](https://discourse.processing.org/t/another-neural-network-trainer/22624/1 "2020-07-14T21:53:09Z")

</div>

In case someone needs it - [https://github.com/iiliev/processing\_neural\_net\_trainer](https://github.com/iiliev/processing_neural_net_trainer)  
Cheers…

---

<div class="post-metadata">

**Author:** ![jeremydouglass](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.processing.org/jeremydouglass/32/20_2.png) [@jeremydouglass](https://discourse.processing.org/u/jeremydouglass)\
**Post date:** [July 27, 2020, 6:53pm UTC](https://discourse.processing.org/t/another-neural-network-trainer/22624/2 "2020-07-27T18:53:35Z")

</div>

Thanks so much for sharing this, @ivaylo!

When you say “another” were there other specific neural network options that you were comparing this to, and was there a reason that you made this one instead of using those other options? (as a learning project, or for some aspect of features or usability…?)

---

<div class="post-metadata">

**Author:** ![ivaylo](https://avatars.discourse-cdn.com/v4/letter/i/5fc32e/32.png) [@ivaylo](https://discourse.processing.org/u/ivaylo)\
**Post date:** [August 1, 2020, 1:58pm UTC](https://discourse.processing.org/t/another-neural-network-trainer/22624/3 "2020-08-01T13:58:13Z")

</div>

The ones I found were either:

- incomplete (someone tried to port a JS or a Java one, got stuck and dumped the code hoping someone would help) OR
- relying on esoteric Java libraries probably prohibitive to use/setup in Processing (esp. for beginners) OR
- limited to one hidden layer or a single activation function, etc.
- useful for a specific use case or with limited functionality (say no backward propagation implemented)

Also the more people learn/use NNs in Processing the more having a separate trainer would be beneficial. If there is enough interest I’ll make a proper Processing library out of this. Thanks for the interest…