Skip to main content
Loss of provenance

A previous release had provenance metadata, but this release does not. View details.

numdiff-research

Adaptive Local Polynomial Regression Differentiation (ALPRD) — an ongoing, honesty-first research project on numerical differentiation of unevenly sampled, noisy 1-D data.

Status: Phase 4 concluded at V2.2 (estimator frozen at the V2.0 formula; V2.2 added physical-sensor validation, a completed WLS-path oracle inequality, two falsified-and-reverted fixes, a documented scope boundary, and the paper draft) (evidence-based stopping condition: all remaining improvements require external literature, external code/data, or genuinely new mathematics — see the V1.4 log section). Current self-assessed publication readiness: 7.5/10. Claims are strictly separated into Proven / Derived / Tested / Hypothesized, and negative results are kept in the record (V1.0's outlier-robustness claim was retracted after a control experiment; V1.0-V1.2's noise estimator was found miscalibrated by 2.2x and fixed in V1.3; two V1.4 hypotheses were tested, falsified, and recorded).

Layout

docs/RESEARCH_LOG.md      the full record: derivations, assumptions, benchmarks,
                          failure analyses, version history (V1.0, V1.1, ...)
src/methods.py            ALPRD variants + all baseline methods
src/datasets.py           benchmark dataset generators (exact ground-truth
                          derivatives, incl. a Lorenz trajectory via solve_ivp)
src/benchmark.py          the full 35-scenario suite; writes results/*.csv
tests/test_methods.py     correctness + equivalence tests (plain python, no
                          pytest required)
results/                  raw per-run CSVs and scaling measurements — every
                          number cited in the log traces to a file here

Reproduce

pip install numpy scipy pandas
python tests/test_methods.py     # 5 tests, all must pass
python src/benchmark.py          # ~2 min; rewrites results/results_v1_1.csv

Everything is seeded (seed=12345 throughout); re-running the suite must reproduce the CSVs bit-for-bit on the same numpy/scipy versions (numpy 2.5.1 / scipy 1.18.0 were used for the committed results).

Method in one paragraph

At each query point, ALPRD fits a weighted local polynomial (tricube kernel, degree 2) over a k-nearest-neighbor window of the actual sample times — no resampling to a uniform grid — and reads the derivative off the linear coefficient, together with a finite-sample variance estimate. V1.1 selects k per point by minimizing an estimated Bias² + Variance (plug-in rule built from the derived leading-order bias and exact variance functionals). V1.2 makes every fit Tukey-biweight-robust (IRLS), refines the curvature pilot at the selected bandwidth, and vectorizes the whole selector; under 3% gross contamination its error is now ~16x lower than any non-robust method in the suite while remaining near-best on clean and noisy data. Known open problems, failure modes, and the improvement queue are tracked in docs/RESEARCH_LOG.md Steps 8–13 of each version.

Quick Start

import numpy as np
from alprd import robust_differentiation

t = np.sort(np.random.uniform(0, 10, 100))
y = np.sin(t) + np.random.normal(0, 0.1, 100)
derivative = robust_differentiation(t, y)

Metadata

Release files for alprd 2.4.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Loss of provenance

These files do not have provenance metadata, despite files in a prior release including this data.

We recommend further investigation before installing these files. View details.

Source distribution (sdist)

Source distribution for alprd 2.4.0
File Size Uploaded Provenance Actions
alprd-2.4.0.tar.gz 35.2 kB Without provenance Details for alprd-2.4.0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for alprd 2.4.0
File Interpreter ABI Platform Actions
alprd-2.4.0-py3-none-any.whl Python 3 none any Details for alprd-2.4.0-py3-none-any.whl

Total release size: 66.0 kB

Release files / alprd-2.4.0.tar.gz

Loss of provenance

This file belongs to a release without provenance metadata, despite a prior release including this data.

We recommend further investigation before installing this file. View details.

Download URL alprd-2.4.0.tar.gz
Size 35.2 kB
Tags Source
SHA-256 checksum
How to use checksums
83df0389cdbaec35bf3a9e228b0a32b9edc6201d835228b11838d65e31579b30
BLAKE2b-256 checksum
How to use checksums
620acfd24716a4d4ed29af0b0c48370d2a477399eb3bb86ffd7fb721fec3056b
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.12.10

Release files / alprd-2.4.0-py3-none-any.whl

Loss of provenance

This file belongs to a release without provenance metadata, despite a prior release including this data.

We recommend further investigation before installing this file. View details.

Download URL alprd-2.4.0-py3-none-any.whl
Size 30.7 kB
Tags Python 3
SHA-256 checksum
How to use checksums
4012b5e4a400a3bf2b5d7f6a57d78442153aa4e1533bcaed6883a68311167aa6
BLAKE2b-256 checksum
How to use checksums
c56a94e1bf4ac1f2047ace65bc78d68f77029e751e4a84f9a5f6d8b707cfc85b
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.12.10

Security for alprd 2.4.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Loss of provenance

A previous release of these files had provenance metadata, but this release does not.

While this could reflect a routine change by the project maintainers, it may also indicate that the release files have been compromised.

Release comparison

  • Removed 2.2.0 2 files With provenance
  • Added 2.4.0 2 files Without provenance

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed

This release

2.4.0 This release

2 release files Loss of provenance

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page