I wonder how they manage to train a deep network using single CPU cores and no GPU... without taking months to finish the process.
They use liberties in their definition of "deep", I suspect
I wonder how they manage to train a deep network using single CPU cores and no GPU... without taking months to finish the process.
Of the 100s there seems to be all sorts of indicators buying and selling. It is quite disappointing that none work. However I expect the big money boys have bought up any that do. Only a matter of time though and whammmmo....... I'm in.
I have designed 3 systems so far mainly based on standard indicators but alas poor results so far. I have yet to optimise the 3rd. I employed a 3rd party to code them up. No Martingale etc. as too risky and expensive.
Like what ?It's very debatable whether Moving average, bollinger bands and all other technical indicators actually do work. I am not very fond of them myself. Maybe you should try another approach?
Like what ?
If someone could take the software developed by PAL and make it multi-threaded could be a hit. It's now integral part of my trading but running different instances on different cores/PCs is not ideal way of doing things from me. The developer is probably satisfied and not planning any changes. Then I could use command line execution and an API to IB for set number of stocks that satisfy scan settings. I'm sure others would pay top dollar for these features.
Alternatively any software like PAL that can scan 1000s of stocks and suggest a long and a short portfolio would be of high value and interest. Not simple TA but something like machine learning or deep neural networks.
It is not easy to quantify fundamentals.Maybe adding fundamentals to the price movement?