Feature engineering

Feature-engineering and selection research tools: multi-resolution feature stacking (multi_resolution), regime-adaptive windows (adaptive_roll / adaptive_volatility), a Granger-causality test (granger_causality) and incremental moments (IncrementalMoments).

multi_resolution(func, X, windows, **kwargs)	Stack a window-based feature computed at several resolutions.
adaptive_roll(X, func, windows, regimes, ...)	Apply a window-based feature with a regime-dependent window.
adaptive_volatility(X, windows, regimes[, ...])	Regime-adaptive realized volatility (worked example of adaptive_roll).
granger_causality(x, y[, lag])	Granger-causality F-test: does x help predict y?
IncrementalMoments()	Online mean / variance via Welford's algorithm (O(1) per update).