Title: Tensor decompositions reveal neural signatures of cognitive and behavioral state changes
Abstract: Perceptions, thoughts, and actions unfold over millisecond timescales, while learned behaviors can require many days to mature. While experimental advances in neuroscience enable large-scale and long-term recordings with high temporal fidelity, extracting interpretable descriptions of how rapid single-trial circuit dynamics change slowly over many trials remains challenging. I will show how a well-known tensor decomposition model (canonical polyadic decomposition / PARAFAC) can meet this challenge by extracting three interconnected, low-dimensional descriptions of neural data: neuron factors, reflecting cell assemblies; temporal factors, reflecting rapid circuit dynamics; and trial factors, describing both long-term learning and trial-to-trial changes in cognitive state. Since our initial publication of this approach in 2018, tensor decomposition has become widely adopted in systems neuroscience. My talk will provide several examples of the model’s subsequent use.