Mapping Latent Neural Dynamics Using Recursive Informational Curvature: A Benchmark Across ECoG and EEG

This project introduces Recursive Informational Curvature (RIC), a novel information-geometric metric designed to uncover hidden structure, state transitions, and complexity in neural time series data. RIC quantifies the interplay between signal entropy and recursive feedback, offering a powerful, low-dimensional descriptor of brain state dynamics. We benchmark RIC across both macaque electrocorticography (ECoG) and human resting-state EEG, demonstrating its ability to reveal individual differences, enhance unsupervised clustering, and detect temporal structure that remains invisible to traditional spectral and amplitude-based features. All analyses are performed with fully open-source, reproducible Python code. Key Features: End-to-end Python pipeline for RIC extraction and analysis Application to open-access macaque ECoG and human EEG datasets Benchmarking against classical spectral, complexity, and clustering metrics Tools for unsupervised clustering, surrogate testing, anomaly detection, and visualization Open code and detailed documentation supporting transparency and reproducibility Intended for researchers in computational neuroscience, neuroimaging, and complexity science, RIC provides a new analytic lens for investigating latent neural dynamics, integration, and brain-state transitions.