Statistical physics of large-scale neural activity with loops

Area of Science:

• Computational Neuroscience
• Statistical Physics
• Systems Neuroscience

Background:

• Analyzing large-scale neural activity is crucial for understanding brain function.
• Existing models struggle with the inherent feedback loops in neural circuitry.
• Statistical physics offers a framework for understanding collective neural activity.

Purpose of the Study:

• To develop a scalable statistical physics framework for analyzing large neural populations with feedback loops.
• To enable direct access to information-theoretic measures in complex neural networks.
• To identify optimal neural correlation networks that maximize information content.

Main Methods:

• Developed an exact solution to the maximum entropy problem for networks with feedback loops.
• Applied the framework to large-scale recordings (approx. 10,000 neurons) from the mouse visual system.
• Compared the new framework's performance against existing methods for loop-free networks.

Main Results:

• The new framework accurately models large neural populations with feedback loops, outperforming existing methods.
• It provides direct access to information-theoretic measures previously inaccessible at large scales.
• Models performed better during visual stimulation than spontaneous activity, indicating consistent neural circuitry.

Conclusions:

• The developed framework offers an optimized approach for studying the statistical physics of large neural populations.
• It provides a better description of population activity and reveals consistent neural circuitry.
• The methods have potential applications beyond neuroscience, in other biological networks.