The goal of this NeuroNex Theory team is to develop a new lens — a set of mathematical and computational tools — for interpreting the recordings that are becoming available. These techniques will allow us to deduce when and how observed neurons interact, how these interactions are altered by stimuli, and how they may govern behavior.
Key Research Resources Being Developed and Disseminated
The team will develop statistical tools that will provide insights into the computations performed by neuronal ensembles. They will validate the methods using simulated neuronal networks, and apply them to recordings from the mouse brain. The team will also develop and test their approach using modern deep neural networks that achieve or exceed human performance in many hard tasks. This will allow them to quantify how interactions in these artificial networks change with the stimulus, and compare and contrast the results with data from animals. This broad set of goals requires a combination of approaches and diverse expertise, and the team therefore consists of experimental and theoretical neuroscientists, mathematicians, and statisticians.
Key research resources:1) A new classes of graphical models to study interactions (functional connectivity) between neurons from large-scale recordings via calcium imaging. The new models will address a number of challenges including the non-Gaussianity of spiking activity, non-stationary time series, graph inference from non-simultaneous recordings, and account for latent activity and brain states. 2) Software for large-scale detailed spiking network simulations. The models will be constructed to be biologically realistic and to reproduce the statistics observed experimentally. 3) Pipelines to analyze two- and three-photon imaging data from mouse visual cortex data with a goal to infer interactions between neurons and interactions between neural activity and latent variables in natural scenes. 4) Machine learning models for solving naturalistic tasks that will serve as a testbed for the statistical methods. These models will classify images and videos based on feedforward and recurrent artificial neural networks. 5) Graphical models that augment conventional second-order models, and introduce higher-order interactions to accommodate contextual modulation of neural interactions.