dividing by len(sources[idx])

[carloalbertobono]

Hello,

I am really fascinated by the concept, and trying to apply it in emergency scenarios (social media).

I can't understand one detail of the implementation (377 in Dshap.py):

DataShapley/DShap.py

Line 377 in 303d91d

marginal_contribs[sources[idx]] /= len(sources[idx])

How dividing by the length of an index is beneficial?
Was it meant to support multiple indexes and then the implementation changed?

Sorry if I am missing something obvious

Thanks for the work!