search

ZhengyaoJiang / PGPortfolio

PGPortfolio: Policy Gradient Portfolio, the source code of "A Deep Reinforcement Learning Framework for the Financial Portfolio Management Problem"(https://arxiv.org/pdf/1706.10059.pdf).
GNU General Public License v3.0
1.75k stars 753 forks source link

Trading agent #11

Closed AhmMontasser closed 7 years ago

AhmMontasser commented 7 years ago

Hello great team,

is there a trading agent already built to trade live using the saved trained model ?

I would be interested to help either by building or by improving existing.

Regards,

ZhengyaoJiang commented 7 years ago

Greetings, Yes, our team is doing live-trading now. Of course, in long term, we need more engineers and expert of quantitative trading/ machine learning to expand our business. While currently, we do not eager to recruit new members since the live-trading framework has already been built and tested for quite a while. If you want to join us in future(maybe several months later), you could try to contribute to this open source project first to get familiar with the framework.

Regards Zhengyao

AhmMontasser commented 6 years ago

Thank you

huminpurin commented 6 years ago

I'm also trying to use the trained model for live trading. the problem is i cant find which part of program generates the actions of buying/selling signal. Any help would be appreciated. Thanks!

ZhengyaoJiang commented 6 years ago

cant find which part of program generates the actions of buying/selling signal

Basically, the nnagent class output the portfolio weights. You can subtract two portfolio vector to get the transaction vector.