rLifting 1.0.0
New features
- Native irregular-grid support: pass a sorted
t vector to denoise_signal_offline(),
denoise_signal_causal(), or t_val per sample
to the stream closure from new_wavelet_stream(). The
predict step uses Lagrange interpolation conditioned on the actual
sample positions — no separate regularisation required.
- Two new boundary extensions:
local_linear (OLS extrapolation with configurable
neighbourhood ll_k) and one_sided
(renormalised filter at the edge), bringing the total to five alongside
the existing symmetric, periodic, and
zero.
- SureShrink threshold rule
(
threshold_method = "sure"): per-level Stein Unbiased Risk
Estimator, complementing the existing universal (VisuShrink) rule.
- SCAD shrinkage (
shrinkage = "scad"):
Smoothly Clipped Absolute Deviation (Antoniadis & Fan, 2001), added
alongside the existing hard, soft, and semisoft shrinkages.
tune_alpha_beta(): automatic selection
of the recursive threshold parameters α and β by minimising SURE over
the signal.
diagnose_wavelet(): verification suite
for custom wavelets — perfect reconstruction, vanishing moments,
orthogonality, compact support, and shift sensitivity.
- Seven benchmark datasets bundled:
benchmark_rlifting,
benchmark_wavethresh, benchmark_adlift,
benchmark_nlt, benchmark_rlifting_irregular,
benchmark_adlift_irregular,
benchmark_nlt_irregular — each based on 1,000 Monte Carlo
replications per configuration.
- Eight vignettes: introduction, thresholding and tuning,
causal/stream, boundary modes, irregular grids, extensions, full
cross-package benchmarks, and a real-world case study (BabyECG infant
cardiac monitoring).
Deprecated
- The
method argument in
denoise_signal_offline() and
denoise_signal_causal() is deprecated in favour of the
separate threshold_method and shrinkage
arguments. Old code continues to work via a backward-compatibility shim;
method will be removed in a future version.
rLifting 0.9.0
- Initial release of the package.
- Implements high-performance Wavelet Lifting Transforms.
- Features:
- Unified offline (batch) denoising via
denoise_signal_offline().
- Causal (sliding-window) filtering via
denoise_signal_causal().
- Sample-by-sample stream processing via
new_wavelet_stream().
- Adaptive recursive thresholding (universal rule, hard/soft/semisoft
shrinkage).
- Zero-allocation ‘C++’ core via ‘Rcpp’.
- Six built-in wavelets: Haar, DB2, CDF 5/3, CDF 9/7, DD4, Lazy.
- Three boundary extensions:
symmetric,
periodic, zero.