missing function 'min_max_rescale' in portfolio_learning_strategies.py

[mthede]

Due to missing 'min_max_rescale' function in utils.py, an import error is thrown in portfolio_learning_strategies.py

assume/assume/strategies/portfolio_learning_strategies.py

Line 14 in d48b152

from assume.common.utils import min_max_rescale, min_max_scale

@chiara-fb:

• could you add the missing function to utils.py?
• and maybe add a test for the RL portfolio strategy

Thanks!

[maurerle]

The missing function is found here: https://github.com/chiara-fb/assume/blob/c7756ab2716ed8fe9e362f3403e999a1d4f9a97a/assume/common/utils.py#L762

[chiara-fb]

Noted, will work on a test function next week. I did not push my version of utils with the last update.

[mthede]

Thanks, Chiara! I took the initiative and drafted some tests in the attached PR from Florian. I wanted to have a more detailed look into your portfolio learning strategy anyway :)
Maybe you could adjust the precise values (forecasts, costs, etc.) to test against actual values following your intended reward/bid formulation?

And I proposed a more generalistic version of the min_max_scale function, that can now be used for "both ways" - scale and rescale - by just switching the bounds.

[chiara-fb]

I like your suggestion for the two-way min_max_scaling, Marie! A side note (also in the code review).

Consider substituting line 453 of EnergyLearningStrategy from:
bid_prices = actions * self.max_bid_price
to:
bid_prices = min_max_scale(actions, -1,1, self.min_bid_price, self.max_bid_price)

to allow for more flexibility (i.e. min bid prices different than -1 * max_bid_price).