Decision making and perceptual bistability in spike-based neuromorphic VLSI systems

Abstract

Understanding how to reproduce robust and reliable decision making behavior in neuromorphic systems can be useful for developing information processing architectures in subthreshold analog circuits as well as future emerging nano-technologies, that comprise inhomogeneous and unreliable components. To this end, we explore the computational properties of a recurrent neural network, implemented in a custom mixed signal analog/digital neuromorphic chip, for realizing perceptual decision-making, bi-stable perception, and working memory. The chip comprises conductance-based integrate-and-fire neurons and configurable synapses with realistic dynamics. These circuits are configured to implement a recurrent neural network, composed of excitatory and inhibitory pools of silicon neurons coupled with local excitation and global inhibition. We show how the interplay between excitation and inhibition produces competitive winner-take-all dynamics, which is a feature of decision-making and persistent activity models, and demonstrate that the system generates reliable dynamics capable of reproducing both neurophysiological data and psycho-physical performances in coding and collective distributed computation.978-1-4799-8391-9/15/$31.00 ©2015 IEEE

Neuromorphic Implementation of a Continuous Attractor Neural Network With Various Synaptic Dynamics

“…As mentioned in the introduction section, several neuromorphic circuits and hardware related to the CANN have been proposed. Regarding functionality, Corradi and his collaborators implemented decision making and perceptual bistability in a neuromorphic VLSI chip [53]. However, slow ramp-up activities, which are critical for evidence accumula-tion during decision tasks, cannot be reproduced due to a lack of various synaptic dynamics, especially slow and nonlinear synapses.…”

A Neuromorphic Computational Primitive for Robust Context-Dependent Decision Making and Context-Dependent Stochastic Computation

“…However, the specific network architecture through which neurons perceive the conflicts and reach the maximum consistency solution is still unknown. Several software and hardware models have been proposed to explain how the conflicts are avoided and how the stable state is approached, using attractor dynamics and competitive Winner-Take-All (WTA) networks [6], [7], [8], [9], [10], [11], [12], [13]. However, most of these models are based on the assumption that the network can get out of local minima thanks to external stochastic stimuli or injected noise currents.…”

A Dynamic Neural Network Designed Using Analytical Methods Produces Dynamic Control Properties Similar to an Analogous Classical Controller

“…These chips effectively model neurons and synapses in a compact architecture that consumes orders of magnitude less power than a comparable digital system [1], [2]. Most synthetic neural research has been focused around pattern recognition, image processing, or decision making [3]- [5]. Almost all of these systems do not require quick reactions to external changes or interaction with an unpredictable environment.…”