Move var_total and sumsq implementations to corresponding Data classes

ExaScience / smurff

Bayesian Factorization with Side Information in C++ with Python wrapper

MIT License

70 stars 14 forks source link

Move var_total and sumsq implementations to corresponding Data classes #57

Closed motoharu-yano closed 6 years ago

motoharu-yano commented 6 years ago

Currently everything is implemented in MatrixDataTempl using template specialization. Even though methods are defined as virtual in Data class. We should move implementations to DenseMatrixData, SparseMatrixData and ScarseMatrixData. Also SparseMatrixData and ScarseMatrixData should NOT share the implementation. Current implementation using Eigen Sparse matrix does not consider implicit zeroes, so it is invalid for SparseMatrixData case.

motoharu-yano commented 6 years ago

Done in https://github.com/ExaScience/smurff/commit/c0bcb1ceee2972f6369b2e5e333760af2c27f72f