State-Space Kinetic Ising Model Code and Data for Neuronal Spiking Analysis

Overview:

This deposition provides the complete Python code and associated large-scale data used for analyzing non-stationary and nonequilibrium neuronal spiking activity based on a state-space kinetic Ising model. The model captures time-varying firing rates and causal couplings, allowing the estimation of time-asymmetric dynamics such as entropy flow.

Key Features:

• Comprehensive Analysis Tools:
  Implements a state-space kinetic Ising model to evaluate neural spiking data, capturing dynamic changes in firing rates and causal interactions.

• Reproducible Environment:
  Tested on Python 3.8.10 and compatible with Python 3.8+. Utilizes popular libraries including numpy, matplotlib, and scipy; joblib or numba are optionally used to speed up computations.

• Open Science & Data Management:
  Due to the large volume of data (with some files exceeding several gigabytes), the repository leverages Zenodo for data archiving and long-term preservation. GitHub releases are automatically archived on Zenodo and are accompanied by a DOI, ensuring proper citation.

How to Reproduce:

• Figures 1 and 2: Run python main_kinetic/fig1_2.py. This script handles parameter loading or generation, EM fitting, and plotting.

• Figure 4: Run python main_kinetic/fig4.py, which similarly configures parameters, loads or generates data, and produces Figure 4.

Additional Details:

• By default, the scripts generate all data and parameters from scratch. Options to load precomputed data will be supported in future updates.

• The repository includes utility scripts for synthesizing spike data, parameter estimation via an EM algorithm, and routines for comparing different methods of entropy flow estimation.

For further details, please refer to the full documentation in the GitHub README and our preprint: https://arxiv.org/abs/2502.15440.