Extracting feature importance

[bschilder]

Hello!,

How would I go about extracting feature importance scores from a trained GenoCAE model? Some methods I'd like to try out:

1. Simply extract feature weights
2. Compute SHAP scores (or something akin to it)

I think part of my issue is not being sure how to recontruct the model from the GenoCAE outputs, since I see the weights are all stored in weights subfolder but not sure how to import them.

As a side note, I'm more familiar with the format where the entire model (architecture, weights) is saved as one .h5 file. Is there a way to save GenoCAE models in this way?

Thanks so much!,
Brian