Ulises Pereira-Obilinovic
Scientist, Allen Institute for Neural Dynamics
Dynamics in Networks with Learning Rules Inferred from Data
HSB G328, January 17, 1:30 PM
https://washington.zoom.us/j/99088001918
Attractor networks are an influential theory for memory storage in brain systems. However, this theory has recently been challenged by the observation of strong temporal variability in neuronal recordings during memory tasks. First, I will present a study of a recurrent network model in which both learning rules and the distribution of stored patterns are inferred from distributions of visual responses for novel and familiar images in the inferior temporal cortex (ITC) (Pereira & Brunel, 2018, Neuron). We show that there exist two types of retrieval states: one in which firing rates are constant in time and another in which firing rates fluctuate chaotically. We develop a dynamical mean field theory to analyze the network dynamics and compare the theory with simulations of large networks. In the online learning scenario in which the network learns and forgets continuously we show that for a forgetting timescale that optimizes storage capacity, the qualitative features of the network’s memory retrieval dynamics are age-dependent: most recent memories are retrieved as fixed-point attractors while older memories are retrieved as chaotic attractors characterized by strong heterogeneity and temporal fluctuations (Pereira-Obilinovic, Aljadeff, Brunel, 2023, PRX). When these learning rules are temporally asymmetric, they transform a sequence of random input patterns into synaptic weight updates. After learning, recalled sequential activity is reflected in the transient correlation of network activity with each of the stored input patterns. Using a mean-field theory, we derive a low-dimensional description of the network dynamics and compute the storage capacity of sequences (Gillett, Pereira, Brunel, 2020, PNAS). We found that multiple characteristics of the recalled attractor, chaotic and sequential activity are consistent with experimental observations. If time permits, I would like to give you a glimpse of how we are using similar network models and analytical tools for building multiregional models constrained by anatomy and trained in Neuropixels recordings and behavior.