search

nlintz / TensorFlow-Tutorials

Simple tutorials using Google's TensorFlow Framework
6.01k stars 1.51k forks source link

Time series forecasting using LSTM #38

Closed 26medias closed 8 years ago

26medias commented 8 years ago

Hi,

Thanks for those examples, they are very useful.

If you are still working on more examples, it'd be awesome to have an example of time series forecasting using LSTM. There's a lot of search for this, and no working example so far.

It's mostly the preparation of the input data that is confusing for me.

hunkim commented 8 years ago

I think it's a very good idea. Do you have some data for that?

26medias commented 8 years ago

A simple sin? It's an easy signal, easy to train with low error.

A Keras tutorial actually uses a sin for their example: http://danielhnyk.cz/predicting-sequences-vectors-keras-using-rnn-lstm/

hunkim commented 8 years ago

Sounds good. I'll look into it. BTW, if you can come up with a PR, it would be also wonderful.

26medias commented 8 years ago

PR?

hunkim commented 8 years ago

I meant a pull request. :-)

On Sat, May 28, 2016 at 2:46 PM, Twenty-Six medias, Inc. < notifications@github.com> wrote:

PR?

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/nlintz/TensorFlow-Tutorials/issues/38#issuecomment-222328922, or mute the thread https://github.com/notifications/unsubscribe/AA3DVyCWM6TUew9NdHyPcRVPg6YFVP0sks5qGKmWgaJpZM4Io_eU .

26medias commented 8 years ago

Oh ok :)

I don't have anything to contribute at the moment. I'm still in the transition from JS to Python, my code is not pretty.

I wish there was a good bridge to use TF in NodeJS :)

nlintz commented 8 years ago

Before I started machine learning I was a node hacker so I feel your pain. Karpathy wrote some pretty cool convnet stuff for javascript which I'd encourage you to check out if you have the time. http://cs.stanford.edu/people/karpathy/convnetjs/

26medias commented 8 years ago

I already played with it quite a bit (https://github.com/26medias/Kobol) But it's very slow to train. Being able to use GPUs is much more efficient to test models faster.