search

AlexanderGuillermoSeguraBallesteros / RRL

Recurrent Reinforcement Learning (RRL)This is a repository for the implementations of RRL, mainly following Moody's work, other authors will be given credit as well on the go.
9 stars 4 forks source link

Questions on coding #2

Open hddy2000 opened 6 years ago

hddy2000 commented 6 years ago

Dear Alexander, I'm reading your codes, simultaneously with the paper, and get a question: There is an 'eta' in the paper, that described as the magnitude of influence of the return Rt on the Sharp Ratio St,(in formula 15)but this parameter seems not to be included in your code.I assume you set the eta into 1, did you?

Eddie,

hddy2000 commented 6 years ago

Here's the link to the moody's paper I mentioned, wanna make sure we are on the same page: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.1.7210&rep=rep1&type=pdf

AlexanderGuillermoSeguraBallesteros commented 6 years ago

That's correct, in the code uploaded I don't use the differential Sharpe Ratio. It is implemented as batch learning.

hddy2000 commented 6 years ago

Hi,I actually find out that by using the 'rewards' you are assuming eta to be 0,not 1,which is exactly as formula 16 described:) Have you considered working out a version with eta,and make it tunable? It could be very fun of that!

luckykanwar commented 6 years ago

Hi @hddy2000 and @AlexanderGuillermoSeguraBallesteros,

Its great to find a working code sample for this work.

I am working on a Master's thesis in the area of Deep Reinforcement Learning and wasn't able to get my head around how to implement the RRL algorithm.

I would greatly appreciate if you guys could give me some ideas on how to improve this model.

Thanks. Nitin