addressalign-toparrow-leftarrow-rightbackbellblockcalendarcameraccwcheckchevron-downchevron-leftchevron-rightchevron-small-downchevron-small-leftchevron-small-rightchevron-small-upchevron-upcircle-with-checkcircle-with-crosscircle-with-pluscrossdots-three-verticaleditemptyheartexporteye-with-lineeyefacebookfolderfullheartglobegmailgooglegroupshelp-with-circleimageimagesinstagramlinklocation-pinm-swarmSearchmailmessagesminusmoremuplabelShape 3 + Rectangle 1ShapeoutlookpersonJoin Group on CardStartprice-ribbonShapeShapeShapeShapeImported LayersImported LayersImported Layersshieldstartickettrashtriangle-downtriangle-uptwitteruserwarningyahoo

Stock market prediction with recurrent neural networks

  • Nov 5, 2012 · 7:30 PM
  • This location is shown only to members

Last night's meetup went great, I think everyone enjoyed the talk. I brought up how I'm currently using recurrent neural networks for stock market prediction and practically got heckled out of the room. The speaker also seemed to think recurrent neural networks were rubbish and just don't work.  Based on my own experience, I strongly disagree!

So I'd like to prove everyone wrong by demonstrating how I'm using recurrent neural networks for stock market prediction specifically, and I'd also like to more generally discuss current research into effectively training recurrent neural networks, why people have had trouble with them in the past, why LSTMs suck, and why standard RNNs are so interesting and powerful now.

So come for a cool demo, free beer, and a discussion about some very exciting research!

Join or login to comment.

  • ABDULLAH

    hello i confuse about the formual for recurrent (elman)neural networ plz some can some one sent me the full mathematical formula of elman neural network

    April 22, 2013

  • A former member
    A former member

    While I understand that the basic assumptions made in the talk last night preclude stock market prediction being fruitful (namely fallible, human-like AI) I don't see how that would explain how companies like Target can predict when a person is pregnant before even they know. So I am more of a proof is in the pudding type and I look forward to this meetup.

    October 2, 2012

    • Dave S.

      Yeah, should be pretty interesting, I'm excited to give the talk.

      Re: fallible, human-like AI vs how target can know things about you just by looking at purchase history, it really comes down the the fact that AI doesn't exist and won't for a long time, while machine learning is more about stats than trying to build fallible human intelligence. Machine learning works today, AI doesn't.

      So basically, HAL2000 is a ways off, but self driving cars, creepily specific advertising, and algorithmic trading are very much real things.

      October 3, 2012

  • Praful K.

    Thanks Dave. You should also look up Resonance Capital and DE Shaw. This is one of they many techniques they use, and have been super successful.

    October 2, 2012

    • Dave S.

      That's neat, I know quants have been using LSTM RNNs and neural networks in general for all kinds of stuff. I'm sure some of them are using the method I'll be talking about, but it may not be many... the research is very new, there's a ton of skeptics, and its a bitch to implement correctly. I'll be talking about some of the difficulties with implementation, so hopefully it will be interesting for everyone.

      October 3, 2012

  • Dave S.

    Oh, by the way, if anyone wants to prepare for the discussion ahead of time, these are the applicable papers I'll be referring to alot: http://www.icml-2011.org/papers/532_icmlpaper.pdf and http://www.cs.toronto.edu/~ilya/pubs/2011/LANG-RNN.pdf

    October 2, 2012

61 went

People in this
Meetup are also in:

Sign up

Meetup members, Log in

By clicking "Sign up" or "Sign up using Facebook", you confirm that you accept our Terms of Service & Privacy Policy