To your first question: yes, for some definitions of "successfully"[1][2][3][4]. It's difficult, and it isn't as though you train your algorithm and watch it ma

To your first question: yes, for some definitions of "successfully"[1][2][3][4]. It's difficult, and it isn't as though you train your algorithm and watch it make money. It's more that some functions of data analysis become incrementally easier or superior. It's fundamentally similar to applying RNNs to any other signal processing problem. To your (implied) second question: tech companies like Google, Facebook and Microsoft have a strong incentive to publish their researchers' work. It's good for recruiting and pushing progress forward overall. Financial companies have a strong incentive not to do this, in fact they are incentivized to be secretive. In general this means the only quant finance papers you see are published by academics, not quants. This typically results in fewer papers with stunning results from the financial industry, whereas we're almost getting progress fatigue from the papers published by tech companies (where "stunning" means impressive and maybe immediately useful). [1]: https://arxiv.org/pdf/cond-mat/0304469.pdf https://arxiv.org/pdf/cond-mat/0304469.pdf [2]: http://cs229.stanford.edu/proj2012/BernalFokPidaparthi-FinancialMarketTimeSeriesPredictionwithRecurrentNeural.pdf http://cs229.stanford.edu/proj2012/BernalFokPidaparthi-Finan... [3]: https://clgiles.ist.psu.edu/papers/MLJ-finance.pdf https://clgiles.ist.psu.edu/papers/MLJ-finance.pdf [4]: http://www.kolegija.lt/dokumentai_img/Maknickiene-3-8-Pages-from-IITSBE-2011-2-Nr11.pdf http://www.kolegija.lt/dokumentai_img/Maknickiene-3-8-Pages-...