Computation-through-Dynamics Benchmark: Simulated datasets and quality metrics for dynamical models of neural...

Area of Science:

• Systems Neuroscience
• Computational Neuroscience
• Neural Dynamics

Background:

• Systems neuroscience aims to understand how neural ensembles perform neural computation, transforming inputs into behavior.
• Neural dynamics, describing temporal activity evolution, is a key framework for understanding neural computation.
• Current computational models infer neural dynamics from activity, but lack realistic validation datasets and performance metrics.

Purpose of the Study:

• Introduce the Computation-through-Dynamics Benchmark (CtDB) to address limitations in modeling neural computation.
• Provide synthetic datasets that mirror biological neural circuit properties.
• Establish interpretable metrics for evaluating model performance in inferring neural dynamics.

Main Methods:

• Developed synthetic datasets reflecting fundamental computational properties of neural circuits.
• Created interpretable metrics for quantifying the accuracy of inferred neural dynamics.
• Established a standardized pipeline for training and evaluating computational models.

Main Results:

• The CtDB provides realistic synthetic data for validating neural dynamics models.
• CtDB offers validated metrics for assessing model performance.
• Demonstrated CtDB's utility in guiding model development, tuning, and troubleshooting.

Conclusions:

• The Computation-through-Dynamics Benchmark (CtDB) is a critical platform for advancing neural dynamics modeling.
• CtDB enables better understanding and characterization of neural computation.
• Facilitates the development of more accurate computational models for systems neuroscience research.