Matrix factorization for rectangular array data with very few observations.
This is potentially useful for a variety of situations where you want to model pairwise interactions between entities:
- Game scores between sports teams
- Uses of products by customers (as in a recommender system)
- Interactions between drugs and biological samples (as in a drug discovery setting)
- etc.
WARNING: this package is under development -- its API is almost sure to change!
See Bracketology.jl for example usage of this package in a sports analytics application.
SparseMatFac.jl is intended to be the "sparse analog" of MatFac.jl.