Data

Tools

Indexes

Applications

Experiments

Open Source Science

Recent Releases of https://github.com/cbueth/delaynet

https://github.com/cbueth/delaynet - Statistical Normalisation for Network Analysis

New Features

Network Metric Normalisation - Added z-score normalisation for network analysis metrics against random graph ensembles - All network metrics now support normalise=True parameter to compare against directed Erdos–Rényi $G(n,m)$ null models - Configurable ensemble parameters: n_random (default: 20) and random_seed for reproducibility

Improvements

  • Enhanced statistical rigor for network analysis results
  • Updated dependency: synthatdelays>=1.0.2
  • Added Zenodo DOI badge for improved citation support

Notes

Full Changelog: https://github.com/cbueth/delaynet/compare/v0.3.0...v0.3.1

- Python
Published by cbueth 11 months ago

https://github.com/cbueth/delaynet - delaynet: Reconstruct and analyse delay propagation networks

Delaynet is a Python package for reconstructing and analysing delay functional networks from time series. It offers: - Detrending and data preparation utilities - Multiple connectivity measures with unified p-value output and optimal lag selection - End-to-end network reconstruction and analysis tooling

Documentation: https://delaynet.readthedocs.io/ Repository: https://github.com/cbueth/delaynet Issues: https://github.com/cbueth/delaynet/issues

Installation

bash pip install delaynet

Python 3.11–3.13 are supported.

Quickstart

```python import numpy as np import delaynet as dn

Example data: 5 nodes, 300 time points

rng = np.random.defaultrng(1520) data = rng.standardnormal((300, 5))

Pairwise connectivity (Granger causality) with lag search up to 10

pval, lag = dn.connectivity(data[:, 0], data[:, 1], metric="gc", lag_steps=10)

Reconstruct network (p-value matrix and lag matrix)

weights, lags = dn.reconstructnetwork(data, connectivitymeasure="gc", lag_steps=5) ```

Highlights

  • Unified connectivity and reconstruction workflow

    • Consistent p-value output across connectivity measures and automatic best-lag selection
    • New reconstruct_network function producing p-value and lag matrices for directed delay networks

  • Network analysis module

    • Pruning by statistical significance or multiple-comparison control
    • Core metrics: centralities, link density, reciprocity, transitivity, global efficiency; isolated nodes

  • Performance and robustness

    • Parallel execution support for reconstruction; progress tracking in both sequential and parallel modes
    • High test coverage and cross-platform CI

  • Data and documentation

    • Synthetic data generators (including transportation-oriented scenarios)
    • Structured guides and API reference covering detrending, connectivity, reconstruction, and analysis

Notable changes and compatibility

  • Terminology update: “normalisation” → “detrending” throughout the API and docs
  • Entropy-based connectivity API simplified (direct keyword arguments; deprecated kwargs dicts removed)
  • Symbolization submodule removed
  • Requires Python ≥ 3.11

Links

  • Documentation: https://delaynet.readthedocs.io/
  • Changelog: https://delaynet.readthedocs.io/en/latest/changelog/
  • Issues: https://github.com/cbueth/delaynet/issues

License

BSD-3-Clause

- Python
Published by cbueth 11 months ago

https://github.com/cbueth/delaynet - Stability and Enhancement Update v0.2.0

What's Changed

Version 0.2.0: Major Enhancements and Bug Fixes by @cbueth in https://github.com/cbueth/delaynet/pull/21

  • 🔄 Granger: Rework bidirectional version

  • 🐛 Fix: Stability of random data

    • Remove use of :func:numpy.random.randint()
    • Add test test_gen_rand_data_stability()
    • Add fixed seed to fixture two_fmri_time_series()

  • 🐛 Fix: Fix OS connectivity

    • Rename to fit with US english: synchronisation -> synchronization

  • 📝 EX: Add example comparing connectivities with fMRI data

  • ✏️ Typo: Correct fMRI typo

  • 🐛 Fix: Random time series indexing

  • 📈 Z-Score: Added max_period, exclude current datapoint

    • Added max_periods parameter to Z-Score normalization function to limit the number of periods considered in calculations.
    • Excluded the current point from mean and standard deviation calculations.

  • 📐 Sig: Make time series positional only

  • 🧪 Test: Add automatic tests for all norms and connectivities

    • Uses generated data
    • Approaches not all functioning yet

  • 📚 Add data generation methods

    • Generate fMRI time series
    • Wrapper for all approaches
    • Increased max-args = 8
    • Corrected argument order

  • 📁 Ignore built folder, uses by pip

  • 📚 Doc: Changed setup modality compatible with pip and micromamba

  • 🧪 Tests: Add python 3.10 and 3.12 compatibility

  • 🔄 CI/CD: Change environment caching runner

Full Changelog: https://github.com/cbueth/delaynet/compare/v0.1.0...v0.2.0

- Python
Published by cbueth over 2 years ago

https://github.com/cbueth/delaynet - Release version v0.1.0

What's Changed

  • Merge connectivity decorator and norm checks by @cbueth in https://github.com/cbueth/delaynet/pull/10
  • Generalize Z-Score norm by @cbueth in https://github.com/cbueth/delaynet/pull/11
  • Refactoring, extended norm and connectivity by @cbueth in https://github.com/cbueth/delaynet/pull/13
  • Lint and Doc by @cbueth in https://github.com/cbueth/delaynet/pull/20

Full Changelog: https://github.com/cbueth/delaynet/commits/v0.1.0

- Python
Published by cbueth over 2 years ago