NEWS / Release Notes

Unreleased

[0.5.0]

Breaking changes.

[CHANGED] Default case now does NOT attack cases with same characterisation but different outcome.
[FIXED] Cases incoherent with default cases were not being added in cautious aacbr.
[FIXED] Issues with cases with same characterisation as default.

[0.4.2]

[FIXED] Bugs with sorting fixed, solving issues with remove_spikes not working properly, among others.

[0.4.1]

[FIXED] When the default case is passed as argument and has a different outcome than originally set in the Aacbr instance as default outcome, now this is taken intou account, respecting the explicitly sent default case.

[0.4.0]

API is kept, but massive speed-ups.

[CHANGED] Fitting, for both cautious and non-cautious, is now much faster. In the slowest scenarios in benchmarking contained in tests/, 85% speed-up in non-cautious, and 99.89% speed-up in cautious.

[0.3.0]

More compatible with the sklearn API, although still not passing sklearn automated testing.

[CHANGED] X and y arguments can be used in fit instead of casebase and outcomes
[CHANGED] Attributes only (properly) set after training are now defined in fit, not in initialisation.

[0.2.0]

[CHANGED] When the argument factors of Case is a set, it is converted to frozenset instead. This is done in order to guarantee that cases are hashable. If modifying a case, create a new one with different values.
[CHANGED] Hashes for Case instances are now cached via an attribute.
[CHANGED] past_case_attacks now does not check whether its arguments are in the active casebase. This is now assumed, instead of verified.