Hi Alphien and Alphathon team,
Will the testFXCarrySubmission function call payout many times?
If we put the training in the payout, that would mean retraining the model from scratch for many times?
That seems to waste a lot of time to training if a train cost 15mins. Many times of training will cost a few hours to test the "testFXCarrySubmission".
Therefore,can we training outside the payout, and just call the model in the payout? (I noticed this is not allowed in the notebook)
The choice of parameters coming out of your training should be explained in the notebook that you submit and is part of the solution. The portfolio class should allow you to train without fetching data (please look in detail), so evaluating weights and returns without fetching the data should be quick in the portfolio class. If you want to do it outside the portfolio class you can also do so but please document the approach, make sure you create the right 'helper' function and submit them with your solution.
You can train outside the payout and it is shown in the SP500 submission with good SR.ipynb notebook with a simple loop.
I hope that helps. Thanks, Lionel.
Yes, thanks for your quick reply!
It helps, and we will train outside the payout.