Closed davidweichiang closed 3 years ago
So I am leaning towards this notation. I like that it makes examples like Bayes rule symmetric, uses one less operators, and it seems like a feature that it removes the "global" min interpretation which is a sort of a hack in pytorch / numpy. I have become pretty used to the \cdot, but maybe \odot what correct to begin with?
My sense is that natural contraction is not critical to the main use cases and a bit confusing. If we really need it, we could have \odot_{*} to indicate "contract all in axes in scope".
OK. I think it's simpler overall.
Merges elementwise multiplication and contraction operators into a single operator, \odot.
Closes #35.