Reconstructing the functional connectivity of multiple spike trains using Hawkes models

Area of Science:

• Computational Neuroscience
• Network Science
• Statistical Modeling

Background:

• Reconstructing functional connectivity graphs from neural recordings is crucial for understanding brain networks.
• Multivariate Hawkes processes are statistical models used for this purpose but often require extensive data and prior network knowledge.
• Existing methods may also lead to an unrealistic increase in simulated spike counts over time.

Purpose of the Study:

• To present a novel, data-efficient method for reconstructing functional connectivity in neural networks.
• To enable accurate simulation of neural network activity without requiring prior architectural information.
• To provide a robust and stable tool for analyzing spike train recordings.

Main Methods:

• The study employs a specific class of Hawkes processes utilizing least-square estimators and LASSO penalty criteria.
• The method is designed for efficient simulation and inference of network properties.
• It is tested on small networks modeled using the Leaky Integrate and Fire (LIF) framework.

Main Results:

• The proposed method accurately detects both excitatory and inhibitory connections in small neural networks.
• It demonstrates robustness against complexities such as common inputs, weak connections, and chained connections.
• Errors in complex networks can be identified and discarded using objective criteria.

Conclusions:

• This method reconstructs functional connectivity of small networks without prior knowledge or excessive data.
• It offers a significant advantage over existing techniques in terms of data requirements and prior assumptions.
• The method is robust, stable, and suitable for routine use on personal computers for inferring connectivity and generating simulations.