search

lucasmaystre / choix

Inference algorithms for models based on Luce's choice axiom
MIT License
158 stars 27 forks source link

Using features as extra evidence for preference selection #4

Open usptact opened 6 years ago

usptact commented 6 years ago

Thanks for the great package!

Is it possible in choix for each observation (or pair if you wish) to attribute and use some feature vector? The feature vectors, not outcomes alone, will be used to learn the preferences.

Thanks

lucasmaystre commented 6 years ago

Hi @usptact , thanks for your interest! Unfortunately that is not possible with choix at this time.

I'll leave the issue open, maybe at some point I'll get around implementing it.

If you want a simple & quick way to do it, you could use a logistic regression model (without bias) with a "meta" feature vector x = x_{winner} - x_{loser} for each pairwise comparison outcome. (I think the statsmodels package might do the trick)

usptact commented 6 years ago

Thanks for the statsmodels suggestion!