Before v1.0.0 hits CRAN, I'm planning on a very thorough re-write of the internals for maintainability more than anything else (all API changes are being documented in #35). What better time than this to include sparse matrix support? I can't think of one.
The sparse pipeline will start at fill_gaps(), take an S3 method dispatch in create_lagged_df(), and continue through train_model() and predict().
This one will take a bit of thought because I want to select a matrix package and sparse storage format that is performant, low memory, commonly used in fields like geo-spacial analysis or hydrology, and used in other R ML packages.
Before v1.0.0 hits CRAN, I'm planning on a very thorough re-write of the internals for maintainability more than anything else (all API changes are being documented in #35). What better time than this to include sparse matrix support? I can't think of one.
The sparse pipeline will start at
fill_gaps(), take an S3 method dispatch increate_lagged_df(), and continue throughtrain_model()andpredict().This one will take a bit of thought because I want to select a matrix package and sparse storage format that is performant, low memory, commonly used in fields like geo-spacial analysis or hydrology, and used in other
RML packages.