Product
FerroWave
FerroWave provides discrete and continuous wavelet transforms, multi-resolution decomposition, streaming analysis, regime primitives, and formal proof gates around stable transform laws.
FerroWave turns non-stationary series into scale-aware features. It is the signal-analysis layer for denoising, decomposition, streaming transforms, and regime inputs that need deterministic numeric behavior in research and production. Stable transform and signal-processing laws move into formal proof gates where the numerical contract is precise enough to specify.
Why it matters
Market signals are rarely stable at one frequency or time scale. FerroWave gives research and product teams a shared, reproducible way to separate noise, trend, jump behavior, and regime structure before those signals reach a strategy, risk model, or decision-support surface.
Core capabilities
- Discrete wavelet transform (DWT) — Daubechies, symlet, coiflet, and
biorthogonal bases over
f64slices. - Continuous wavelet transform (CWT) — Morlet, Mexican hat, and Paul wavelets for time–frequency localization on tick and bar data.
- Multi-resolution analysis (MRA) — additive decomposition of a signal into approximation + detail bands across levels.
- Regime primitives — change-point statistics and scale-energy ratios for repeatable regime and volatility-state workflows.
- Formal proof gates — machine-checked contracts for stable transform, alignment, normalization, and bridge laws.
When a system needs a denoising filter, a level- detail coefficient, or a wavelet packet, FerroWave provides the common implementation instead of recreating the math in each product.
Examples
Denoise a signal
Use MODWT denoising to reduce micro-noise while preserving jumps.
Streaming CWT
Maintain a rolling continuous-wavelet view of incoming samples.
Jump detection
Detect jumps and discrete events in a return series using wavelet detail coefficients.
Multi-scale variance
Decompose variance across time scales with the MODWT.
Regime / Hurst
Estimate regime state and the Hurst exponent from wavelet scale energy.
Streaming denoise
Denoise a live tick stream with the streaming MODWT denoiser.
Adaptive / EMD
Adaptive decomposition with EMD and CEEMDAN for non-stationary signals.
Guides
Getting started
Add ferro-wave to a Cargo workspace and decompose your first signal.
Conventions
Sign, indexing, length, and padding conventions shared across ferro-wave transforms.
Transforms
The transforms and wavelet families ferro-wave provides, and when to reach for each.
Streaming
Streaming wavelet primitives for per-tick updates at nanosecond latency.
Analysis
Denoising, coherence, and multifractal analysis primitives in ferro-wave.