Repository navigation
Feat/tsfresh parity extractors - #20
Merged
Merged
Conversation
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
mean_second_derivative_central, value_count, energy_ratio_by_chunks, change_quantiles, benford_correlation and query_similarity_count, with config entries (tsfresh's default parameters; query_similarity_count off unless configured), parity cases and fuzzing against tsfresh 0.21.2. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…r_entropy Spectral features from tsfresh v0.21.2 in a new spectral module. One rfft and one Welch PSD per group are cached in GroupContext and shared across the features' parameters. Parity cases from tsfresh's tests plus generated ones; fuzzed against tsfresh with tolerances that scale with the series (rustfft rounds differently from pocketfft). fft_coefficient gives NaN for a series with NaN or +-inf (documented deviation). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Port tsfresh's wavelet features (pywt.cwt with the Mexican hat wavelet, scipy.signal.find_peaks_cwt with tsfresh's Ricker wavelet) as Efficient tier kernels in the new wavelet module. Results equal tsfresh's bit for bit on x86-64: np.convolve's OpenBLAS ddot / small_correlate summation order, numpy's SVML exp and pywt's pow-based wavelet table are reproduced, since tied neighbouring coefficients are decided by rounding. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…t and augmented_dickey_fuller Port tsfresh's statsmodels-based feature calculators (acf, pacf via Levinson-Durbin, AutoReg OLS, adfuller with AIC/BIC/t-stat/None autolag and MacKinnon p-values) to Rust kernels in the Efficient tier, with config sections, parity cases and fuzz wiring against tsfresh. Deliberate deviations: ar_coefficient gives NaN on NaN/inf (tsfresh raises); augmented_dickey_fuller treats exact fits as zero residual variance and breaks AIC/BIC ties from collinear lags towards the smaller lag (statsmodels decides both by rounding noise). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ation Drop the OpenBLAS ddot and SVML exp emulation and the CPU-feature dispatch: one plain direct convolution, the standard exp and t * t, so results are the same on every platform. cwt_coefficients now matches tsfresh within rounding (TOLERANT); number_cwt_peaks can differ from tsfresh where neighbouring coefficients tie in real arithmetic (documented deviation). The fuzz test checks number_cwt_peaks against scipy's peak finder run on the same portable arithmetic. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Skip ar_coefficient / augmented_dickey_fuller fuzz cases whose least squares designs are ill-conditioned (cond > 1e6, which includes rank-deficient designs and collinear-lag AIC/BIC ties) or contain underflowing values; statsmodels' results there are rounding-dependent. Compare ar_coefficient relative to the largest coefficient of its fit. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Add binned_entropy, lempel_ziv_complexity, permutation_entropy, approximate_entropy (Comprehensive tier), friedrich_coefficients and max_langevin_fixed_point, with tsfresh's default parameters, parity tests from tsfresh v0.21.2 and differential fuzzing. - binned_entropy reproduces np.histogram's edges and binning; the shared `binned_entropy` helper is pub(crate) for fourier_entropy. - Deliberate deviations: binned_entropy gives NaN for ±inf (tsfresh raises); permutation_entropy ranks ties by position (tsfresh's tie order depends on numpy's CPU-specific argsort). - The Langevin features reproduce pd.qcut (incl. its interval labels), pandas' group-by means and np.polyfit; roots come from a balanced companion matrix, polished by Newton steps. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Also drop fourier_entropy's private binned_entropy/np.histogram copy in favour of the shared entropy::binned_entropy, and silence statsmodels warnings in the ar_coefficient fuzz comparison. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.