More approximations enabled

outbrain / outrank

A Python library for efficient feature ranking and selection on sparse data sets.

https://dl.acm.org/doi/10.1145/3604915.3610636

BSD 3-Clause "New" or "Revised" License

19 stars 3 forks source link

Closed SkBlaz closed 1 year ago

SkBlaz commented 1 year ago

Namely,

--mi_approximation_factor, a term that enables sub sampled estimation of relational space (this is mostly the bottleneck)
--min_support parametrization of min support for entropy estimators. This enables skipping computation where support for a value is too small. Default is at least 2, but higher values seems to yield similar results (this has to be further evaluated).

Added also printout of arguments for easier orientation.

SkBlaz commented 1 year ago

@miha-jenko @bmramor I'm thinking of closing this PR given the unexpected speed boost that renders it kind of obsolete. Wdyt