From 2dce001581f182e317dcaaab5a80d0a25e3e87b9 Mon Sep 17 00:00:00 2001 From: luizfernandolj Date: Wed, 11 Nov 2026 10:42:29 -0300 Subject: [PATCH] CHANGELOG: log the MLPE change, which is breaking #14 removes the estimator from MLPE: MLPE(LogisticRegression()) now raises TypeError, cv, stratified and shuffle are gone from the signature, and there is no estimator_ after fit. Anyone upgrading without reading the diff finds that at runtime, so it belongs in the file -- under Breaking Changes, which is the section the v0.5.0 entry already uses for exactly this kind of change, and which Unreleased did not have. The entry says plainly that the estimate does not change and that migration is to drop the argument, because the breakage is in the signature and not in the numbers. #13 is not logged: it changed this file's own labels and links, which is not a user-visible change to the library. Co-Authored-By: Claude Opus 5 --- CHANGELOG.md | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+) diff --git a/CHANGELOG.md b/CHANGELOG.md index d6ef82d5..889b503d 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -6,6 +6,25 @@ All notable changes to mlquantify will be documented in this file. ## [Unreleased] +### Breaking Changes + +- **`MLPE` no longer takes an estimator, and is no longer aggregative.** + `MLPE(LogisticRegression())` now raises `TypeError`; `estimator`, `cv`, + `stratified` and `shuffle` are gone from the signature, the remaining + parameters are keyword-only, and there is no `estimator_` after `fit`. + **The estimate does not change**: it was always the training prevalence, + read from the labels. Migration is to drop the argument -- `MLPE()`. + + The one fit that remained was the entire cost of the method. `CC`, which + trains one classifier and nothing else, cost 0.0523 s against `MLPE`'s + 0.0571 s over 142 datasets, and QuaPy's `MLPE`, which is non-aggregative + and counts labels in two lines, cost 0.0001 s -- a ratio of 347x for work + the estimate never used. `fit` now takes 0.55 ms instead of about 57 ms. + Against QuaPy the training-time ratio falls from 347x to 4.99x, training + memory from 6.90x to 2.04x, estimation memory from 1.90x to 1.00x, and + training plus estimation from 8.51x to 1.38x. The whole method over 142 + datasets takes 50 s of CPU, against 40 min. (#14) + ### Fixed - **`HDx` bins raw features on their own scale, and its estimates change.** It