Rust wavelet transform library for production time series.

FerroWave brings the wavelet canon into Rust: classical transforms, adaptive decompositions, streaming variants, and reference-pinned behavior.

Why this needs production signal processing

!

Research code often starts in Python but production systems need a Rust implementation with stable behavior.

!

Wavelet libraries become hard to trust when transform lengths, boundary modes, and reconstruction rules drift.

!

Latency-sensitive systems need streaming transforms without giving up reference-equivalence checks.

How FerroWave handles it

01

Select the transform family

DWT, MODWT, SWT, WPT, CWT, and adaptive decompositions are exposed as explicit workflows.

02

Preserve numerical contracts

Reference pinning keeps behavior aligned with known Python and scientific-computing implementations.

03

Move into production

Rust hot paths support low-latency feature generation and signal-processing pipelines.

What makes this FerroWave-shaped

Public positioning covers 12 transforms and five wavelet families.

The home page states PyWavelets equivalence targets.

Verification and Lean theorem counts are public site surfaces.

Common concerns

Is FerroWave only a DWT implementation?

No. The public site covers classical wavelets, adaptive decompositions, Hilbert-Huang workflows, and streaming variants.

Can this be used outside finance?

Yes. FerroWave is positioned for signal processing and AI/ML feature pipelines as well as quantitative finance.