Recurrent Network Models Of Sequence Generation And Memory

Abstract

Sequential activation of neurons is a common feature of network activity during a variety of behaviors, including working memory and decision making. Previous network models for sequences and memory emphasized specialized architectures in which a principled mechanism is pre-wired into their connectivity. Here we demonstrate that, starting from random connectivity and modifying a small fraction of connections, a largely disordered recur- rent network can produce sequences and implement working memory efficiently. We use this process, called Partial In-Network Training (PINning), to model and match cellular resolution imaging data from the posterior parietal cortex during a virtual memory- guided two-alternative forced-choice task. Analysis of the connectivity reveals that sequences propagate by the cooperation between recurrent synaptic interactions and external inputs, rather than through feedforward or asymmetric connections. Together our results suggest that neural sequences may emerge through learning from largely unstructured network architectures.

60 pages, 6 figures

References in corpus (4)

• Complexity without chaos: Plasticity within random recurrent networks generates robust timing and motor control
• Fast non-negative deconvolution for spike train inference from population calcium imaging
• Stimulus-Dependent Suppression of Chaos in Recurrent Neural Networks
• Bayesian spike inference from calcium imaging data

Cited by in corpus (33)

• The physics of brain network structure, function, and control
• Artificial neural networks for neuroscientists: A primer
• Large-scale neural recordings call for new insights to link brain and behavior
• Supervised Learning in Spiking Neural Networks with FORCE Training
• Diversity of meso-scale architecture in human and non-human connectomes
• full-FORCE: A Target-Based Method for Training Recurrent Networks
• Optimal sequence memory in driven random networks
• Understanding the computation of time using neural network models
• Towards an integration of deep learning and neuroscience
• LFADS - Latent Factor Analysis via Dynamical Systems
• Learning spatiotemporal signals using a recurrent spiking network that discretizes time
• The Backpropagation Algorithm Implemented on Spiking Neuromorphic Hardware
• Universality and individuality in neural dynamics across large populations of recurrent networks
• Episodic memory governs choices: An RNN-based reinforcement learning model for decision-making task
• Randomly connected networks generate emergent selectivity and predict decoding properties of large populations of neurons
• Lattice physics approaches for neural networks
• Slow manifolds in recurrent networks encode working memory efficiently and robustly
• Suppression of chaos in a partially driven recurrent neural network
• Spectrum of non-Hermitian deep-Hebbian neural networks
• Interpreting Deep Learning: The Machine Learning Rorschach Test?
• Improved memory in recurrent neural networks with sequential non-normal dynamics
• Optimized brute-force algorithms for the bifurcation analysis of a spin-glass-like neural network model
• Optimal Trajectories of Brain State Transitions
• Habit learning supported by efficiently controlled network dynamics in naive macaque monkeys
• Deep Continuous Networks
• Harnessing behavioral diversity to understand circuits for cognition
• Invertible generalized synchronization: A putative mechanism for implicit learning in biological and artificial neural systems
• A ghost mechanism: An analytical model of abrupt learning in recurrent networks
• Encoding Sensory and Motor Patterns as Time-Invariant Trajectories in Recurrent Neural Networks
• A geometrical analysis of global stability in trained feedback networks
• LOOPER: Inferring computational algorithms enacted by neuronal population dynamics
• Shaping dynamical neural computations using spatiotemporal constraints
• Probing the relationship between linear dynamical systems and low-rank recurrent neural network models