Changelog¶
v0.2.1 (2026-03-21)¶
Fixed¶
- Benchmarks now include
passthrough=Truevariants — tree-based models show proper performance - Binary encoding data leakage fixed in benchmark script (MinMaxScaler was fitting on test data)
- Visualization module
__init__.pynow properly exportsplot_maya_number,plot_maya_grid,render_maya_text - Fixed
plot_maya_number/plot_maya_gridreturn type handling in notebook examples
Added¶
- "Results at a Glance" section in README and docs homepage with benchmark tables
- Synthetic data fallback in fraud detection notebook (works without Kaggle credentials)
- Matplotlib visualization demos in VFD deep dive notebook
- Benchmark notebook rewritten with 5 encoding strategies × 4 models + MCE temporal analysis
- All 6 example notebooks listed in README
v0.2.0 (2026-03-12)¶
Features¶
- VFDEncoder: passthrough mode — New
passthrough=Trueparameter keeps original input features alongside VFD-encoded features. This improves performance with tree-based models (RandomForest, GradientBoosting, XGBoost) that benefit from both the raw signal and multi-scale VFD decomposition.
Fixes¶
- Fixed
gregorian_to_jdn()usage in example notebooks (was passing 3 args instead of 1) - Fixed Haab' and Long Count output format in notebook examples
- Fixed docs URL in notebook 01
- Removed compromised polyfill.io CDN from MkDocs config
v0.1.0 (2026-03-12)¶
Initial release.
Features¶
- VFDEncoder: Vigesimal Feature Decomposition transformer
- Three component modes: full, lite, bars_dots
- Auto-detection of vigesimal levels
- Normalization support
- Negative number handling (abs_sign, shift, error)
- Float handling (scale, round, integer_part)
- Inverse transform support
- MayaCalendarEncoder: Maya Calendar temporal encoding
- Tzolk'in (260-day sacred calendar) features
- Haab' (365-day solar calendar) features
- Long Count (linear day count) features
- Cyclical sin/cos encoding
- Wayeb' binary flag
- Configurable epoch
- Core functions: to_vigesimal, from_vigesimal, to_bars_dots, maya_decompose
- Calendar functions: Gregorian ↔ JDN ↔ Maya calendar conversions
- Visualization: Text and matplotlib rendering of Maya numbers
- Full scikit-learn compatibility (Pipeline, clone, get_params)
- 118 tests passing across Python 3.9-3.12