Train, evaluate, and optimize implicit feedback-based recommender systems.
Full Changelog: https://github.com/tohtsky/irspack/compare/v0.3.0...v0.3.1
df_to_sparse
function improved. If user_ids
/ item_ids
are provided, DataFrame row with unknown user/item ID will be ignored.**_colname
is renamed to **_column
. See #99.IDMappedRecommender
. Instead, use
ItemIDMapper
for item id <-> sparse matrix column index mappingsIDMapper
for user/item <-> sparse matrix row/column index mappings*Optimizer
class. Recommender performance can be now tuned by **Recommender.tune
class method.src/irspack
layout & always use relative import.Full Changelog: https://github.com/tohtsky/irspack/compare/v0.2.4...v0.3.0
Full Changelog: https://github.com/tohtsky/irspack/compare/v0.2.3...v0.2.4
Full Changelog: https://github.com/tohtsky/irspack/compare/v0.2.2...v0.2.3
Full Changelog: https://github.com/tohtsky/irspack/compare/v0.2.1...v0.2.2
Full Changelog: https://github.com/tohtsky/irspack/compare/v0.1.20...v0.2.0
alpha
is now reparametrized by alpha0 = 1 / alpha
(roughtly) following "Revisiting the Performance of iALS on Item Recommendation Benchmarks"
Full Changelog: https://github.com/tohtsky/irspack/compare/v0.2.0...v0.2.1
Full Changelog: https://github.com/tohtsky/irspack/compare/v0.1.19...v0.1.20
Full Changelog: https://github.com/tohtsky/irspack/compare/v0.1.18...v0.1.19
ceil_n_heldout
option for split functions