Large Deviations Properties of Maximum Entropy Markov Chains from Spike Trains

Area of Science:

• Computational Neuroscience
• Statistical Physics
• Information Theory

Background:

• Neuronal spike trains exhibit complex collective statistics.
• Maximum entropy methods offer a principled approach to model such data.
• Understanding statistical properties is crucial for interpreting neural activity.

Purpose of the Study:

• To introduce maximum entropy Markov chain inference and large deviations theory to computational neuroscientists.
• To analyze the statistical properties of inferred models, including fluctuations in correlations, distinguishability, and irreversibility.
• To provide accessible explanations of complex theoretical concepts and their applications in spike train analysis.

Main Methods:

• Utilizing the thermodynamic formalism to derive maximum entropy Markov chains.
• Applying large deviations theory to analyze accuracy and convergence properties based on sampling size.
• Developing and illustrating methods for studying statistical fluctuations in neural data models.

Main Results:

• Established connections between maximum entropy inference, statistical physics, and large deviations.
• Provided theoretical tools to quantify accuracy and convergence in spike train models.
• Demonstrated the application of these methods to analyze correlations, distinguishability, and irreversibility in neuronal data.

Conclusions:

• Maximum entropy Markov chain inference, combined with large deviations theory, provides a powerful framework for characterizing neuronal spike train statistics.
• The approach offers valuable insights into the collective behavior and information processing capabilities of neural systems.
• Accessible explanations and practical examples facilitate the adoption of these advanced techniques in computational neuroscience research.