Skip to content

CLAUDE.md

Latest commit

 

History

History
104 lines (79 loc) · 2.92 KB

CLAUDE.md

File metadata and controls

104 lines (79 loc) · 2.92 KB

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

About This Project

timecorr is a Python toolbox for computing and exploring correlational structure in timeseries data. The main functionality is built around the timecorr function which computes dynamic correlations from timeseries observations and finds higher-order structure in the data.

Development Commands

Setup

# Install dependencies
pip install -r requirements.txt
# or
make init

Testing

# Run tests using nosetests
nosetests tests
# or
make test

# Run tests using pytest (alternative)
pytest

Documentation

# Build documentation
cd docs
make html

# Build notebooks (for tutorials)
make notebooks

Building/Installation

# Install package in development mode
pip install -e .

# Build distribution packages
python setup.py sdist bdist_wheel

Code Architecture

Core Components

  • timecorr/timecorr.py: Main module containing the timecorr() function - the primary entry point for computing dynamic correlations
  • timecorr/helpers.py: Utility functions including:
    • Weight functions (gaussian_weights, laplace_weights, etc.)
    • Correlation functions (isfc, wisfc, autofc, wcorr)
    • Data formatting and manipulation utilities
    • Dimensionality reduction functions
  • timecorr/simulate.py: Data simulation utilities for generating test datasets

Key Concepts

The timecorr function accepts:

  • data: Single numpy array/dataframe or list of arrays (one per subject)
  • weights_function: How timepoints contribute to correlations (default: gaussian_weights)
  • cfun: Correlation function - isfc (within/across subjects), wisfc (weighted), autofc (within subject only)
  • rfun: Dimensionality reduction function to prevent feature explosion when computing higher-order correlations

Data Flow

  1. Input data formatted as timepoints × features matrices
  2. Weights function determines temporal smoothing
  3. Correlation function computes moment-by-moment correlations
  4. Optional reduction function prevents dimensionality explosion
  5. Output maintains temporal structure with correlational features

Testing Framework

  • Uses nosetests (legacy) and pytest (modern)
  • Tests in tests/ directory
  • Test files: test_timecorr.py, test_helpers.py
  • Uses simulated data from simulate_data() function
  • Current tests are basic - more comprehensive tests needed per TODO comments

Documentation

  • Uses Sphinx for documentation generation
  • Documentation source in docs/ directory
  • Examples in examples/ directory
  • Tutorial notebooks in docs/tutorial/
  • API documentation auto-generated from docstrings

Dependencies

Core dependencies:

  • hypertools>=0.7.0
  • scipy>=1.2.1
  • scikit-learn>=0.19.2
  • numpy, pandas (via hypertools)

Development dependencies:

  • nose (testing)
  • sphinx (documentation)
  • duecredit (citations)