Correlating paper and repo methodology

[BenSolomon]

Greetings -

Thank you for sharing your work. Looking through the code you've shared, it appears that they represent the specific use cases described in your paper. For instance, the classification models provided are specifically IBD and NPC-classifying models. In order to apply your work to different use cases, would you be able to point where one could find code in your repo corresponding to the following:

1. Based on the methods section of your paper, the last step in generating a comprehensive AIR representation is generating h_fusion from the tensor fusion module. Can you share what section of the repo represents that final output?
2. Does the repo contain code to reproduce your specific classifiers (i.e. IBD and NPC)? From the methods section of the paper, the first step in the classifier is generating classification probabilities for each receptor based on p_m = f(r_m,θ). Can you say more about the specifics of the classification layer? Is θ the same as h_fusion from above?

Thank you for your time and work