tabnetics 1.0.0
First stable release. See https://tabnetics.org for documentation and the interactive results browser.
Highlights
- Production-stable: moved from beta to production-stable classification
- 56,107 benchmark runs across 278 pipeline profiles on 63 HDLSS datasets
- Val-21 integrated: 7 new decision-focused profiles (winner-bridge, meta-selector, prefilter, regime variants)
- 36/63 datasets at balanced accuracy >= 0.90, 26 perfect-classification datasets
- 41/21/1 strict-holdout SOTA split (above/within/below) — 98% at or above published ranges
- Import-safety fix: module-level
find_repo_root()no longer crashes on pip install - Cross-campaign technical report with Wilcoxon tests and 22-hypothesis scorecard
Install
pip install tabnetics==1.0.0Full changelog: https://tabnetics.org/#whats-new-in-100