online training

[zingomaster]

Hello

Thanks for the wonderful work, i read your paper and almost studied most of the code. However, i don't get the concept of append_experience and agent train in the rolling_train method
I have some questions if i may ask
1- what is the format of the saved experience and how does it affect the model ?
2- how is that different from training the model directly using self._agent.train() ?
3- is the experience mentioned here the same as the mini-batches mentioned in the paper for online learning section 5.3 for example ?

thanks in advance
Sarah Ahmed

[ZhengyaoJiang]

what is the format of the saved experience

It's just the history price movement saved at self.__globaldata

how is that different from training the model directly using self._agent.train()

The only difference is the rolling train will append new history from the test set to the training set, during backtest.

[zingomaster]

Thank you for your reply

The only difference is the rolling train will append new history from the test set to the training set, during backtest.

I don't understand this part clearly (the experience appending part), where is this process actually happening ? i see the append experience appends index of a training data , but how get next batch returns the last omega ? where and when is the last omega saved ?

[ZhengyaoJiang]

where and when is the last omega saved ?

self.__PVM, PVM is short for portfolio vector memory.

[zingomaster]

Thank you.

I have 1 last question please, sorry for being so much demanding.
what is the online_w in datamatrices.append_experience(self, online_w=None) ? why do we need it ?
I can see it's (number of assets + 1, ) but i don't understand
Regards,

[dexhunter]

what is the online_w in datamatrices.append_experience(self, online_w=None) ? why do we need it ?

I think online_w is mostly for live trading. You can probably ignore that in backtest

[zingomaster]

i know it's for live trading, i meant what is it ? i don't see anything telling what is it either in the paper or in the code

[dexhunter]

basically it's portfolio weight

[zingomaster]

okaaay, thank you a lot :) 💯