Control Analysis and Design for Statistical Models of Spiking Networks.

Area of Science:

• Computational Neuroscience
• Systems Neuroscience
• Statistical Modeling

Background:

• Neuronal network activity is often characterized using statistical models based on firing rates.
• These models generate n-dimensional binary time series, representing spiking patterns.
• Such models can be derived from data or postulated theoretically for spiking networks.

Purpose of the Study:

• To rigorously develop analytical methods for assessing the controllability of statistical spiking models.
• Specifically focusing on the point-process generalized linear model (PPGLM).
• To quantify the ease of inducing desired spiking patterns through extrinsic input signals.

Main Methods:

• Development of a novel set of analytical techniques.
• Application of these analyses to the point-process generalized linear model (PPGLM).
• Quantification of network response to extrinsic input signals for pattern induction.

Main Results:

• Established a framework for assaying the controllability of PPGLMs.
• Provided quantitative measures for the difficulty or ease of inducing specific spiking patterns.
• Demonstrated the utility of the analysis for understanding network dynamics.

Conclusions:

• The developed analyses offer a robust method for characterizing PPGLM controllability.
• This framework supports basic network analysis and informs applications like neurostimulation design.
• Enables a deeper understanding of how external inputs influence neuronal network outputs.