Self training neural net strat
#1
Hi all,

I have been backtesting various neural net strats for use as live trade bots but im not all that knowledgeable about them.. from research online it seems that typicaly NN's will need to be trained on data before they can work to any great effect to solve a problem, and that once they start running live they typicaly dont improve themselves any more without further training.


Could someone please clarify:

- do the NN strats train themselves as they run using some form of backpropogation (slowly optimising themselves as time goes on)?

OR

- do we use the backtesting and optimisation of the strat using tools like gekkoga to find the best params and then run the strat with those params to get optimised results?

and

- if they dont train them selves as they run currently, could we write a NN strat that would? (optimises its own TOML file after each trade for example using data about price/volume/and historical market trends and overtime builds/evolves into a strat that works for all market conditions)

Thanks!
  Reply
#2
i ussgest you to read this post . it is a aperfect post comparing NN using gekko
https://forum.gekko.wizb.it/thread-57766.html
  Reply
#3
(02-02-2019, 08:04 AM)batssmasher Wrote: i ussgest you to read this post . it is a aperfect post comparing NN using gekko
https://forum.gekko.wizb.it/thread-57766.html

I enjoyed reading thanks for sharing!
  Reply


Forum Jump:


Users browsing this thread: