Generative Models of Brain Dynamics

Abstract

This review article gives a high-level overview of the approaches across different scales of organization and levels of abstraction. The studies covered in this paper include fundamental models in computational neuroscience, nonlinear dynamics, data-driven methods, as well as emergent practices. While not all of these models span the intersection of neuroscience, AI, and system dynamics, all of them do or can work in tandem as generative models, which, as we argue, provide superior properties for the analysis of neuroscientific data. We discuss the limitations and unique dynamical traits of brain data and the complementary need for hypothesis- and data-driven modeling. By way of conclusion, we present several hybrid generative models from recent literature in scientific machine learning, which can be efficiently deployed to yield interpretable models of neural dynamics.

Brain-heart interactions in the neurobiology of consciousness

“…Consequently, some methods to infer the directionality of brain-heart interactions have been proposed, such as Granger Causality ( Faes et al, 2015 ; Greco et al, 2019 ), Transfer Entropy ( Catrambone et al, 2021 ), and Conditional Entropy ( Kumar et al, 2020 ). However, these methodologies rely on the measurements of causal modulations without considering the physiological priors, which could emerge from casual and not causal co-varying of brain-heart oscillations, as occurs in machine learning models trained with a pattern-based logic ( Ramezanian-Panahi et al, 2022 ). A proposed solution is the modeling of bidirectional interactions through a generative approach ( Ramezanian-Panahi et al, 2022 ; Ramstead et al, 2022 ), considering the ongoing modulations between brain and cardiac oscillations ( Candia-Rivera et al, 2021b , 2022a ), in which two physiologically-inspired models of synthetic ECG and EEG series are coupled considering their mutual influences in the ongoing oscillations at different latencies, on the basis of a generative signal.…”

Phase relations of interneuronal activity relative to theta rhythm

“…Several attempts have been made in the literature to optimize networks of spiking neurons (Lee et al, 2016 ; Wunderlich and Pehle, 2021 ; Zenke and Vogels, 2021 ; Ramezanian-Panahi et al, 2022 ). However, these studies are aimed at finding practical applications in the field of data analysis not brain modeling.…”

Deep social neuroscience: the promise and peril of using artificial neural networks to study the social brain

“…For example, autoencoders could be used to compress neuroimaging data ( Makkie et al ., 2019 ). Unsupervised generative models of brain data could also be used to produce large synthetic datasets for algorithm development or in silico experimentation ( Ramezanian-Panahi et al ., 2022 ; Jain et al ., 2023 ).…”