is there any backtesting with your machine-learning algos that can generate some profit? complex quantitative modeling is one thing, making profit is another thing. Lots of times a complex model is not necessarily a good model.
Check out pages 52-53 of the whitepaper.