Thomas Wagenaar

Please visit the wiki before asking these kinds of questions (or other issues asking the same thing). See how to save networks and load them [here](https://github.com/cazala/synaptic/wiki/Networks#tojsonfromjson).

You might want to take a look at [Neataptic](https://github.com/wagenaartje/neataptic), it has some more LSTM examples. It also offers a `clear()` training option which basically tells the network to not remembering...

Seems to be working fine for me. [JSFiddle](https://jsfiddle.net/wagenaartje/wmeq2mLh/)

Maybe you will have some more success with [Neataptic](https://github.com/wagenaartje/neataptic/wiki/Hopfield), the front-end is very similar to Synaptic but the back-end is rather different.

Was mentioned here too - https://github.com/cazala/synaptic/issues/26.

@AlexisTM I wrote a JSFiddle of the demo that uses `requestAnimationFrame` instead of `setTimeout`, problem still occured. https://jsfiddle.net/wagenaartje/fuh36aex/1/. I also set the setTimeout delay to 1000 and it worked fine...

Nice catch! I used `setOptimize(false)` and it worked again. https://jsfiddle.net/wagenaartje/4516xtze/7/ I also noticed that the perceptron has way too many neurons in its hidden layer. Could easily be reduced. Just...

What code are you exactly running? We can't help you if we don't no what is going wrong...

Seems to working fine here... [jsfiddle]( https://jsfiddle.net/buvwzxnz/)

I have actually created a PR yesterday that removes all those incorrect loops. So stay tuned :)