Optimazing weights. #28
Description
Sorry, to bother you. It is not an issue but a matter of understanding on my part.
First of all. Thanks a lot for the example. Great job. I am trying to derive my own strategy based on the example.
Before basing my way forward on a false understanding of the strategy, I would like to ask the following question:
The weights w0-w8 are hyperoptable (at least it seems to me like that). If I am not mistaken, those are also used to derive the target score.
Ignoring the fact, that the sample parameter aren't aggregating to one (only the default ones), I am wondering what this implies for the strategy.
Lets assume, we want to hyperopt the parameters.
- The model trains with the sample parameters given in the strategy.
- In a second step, hyperopt tries to optimize said parameters, without retraining the model obviously.
-> This is likely not going to work.
My question is now the following: Did I understand this correctly? Especially how the target score to train the model is generated or did my train of thought derail completely?
Best regards and thanks a lot.