Hyper-Flexible Neural Networks: Rapidly Switching between Logic Operations in a Compact 4-Neuron Circuit

Abstract

AbstractNeural networks of various levels exhibit rapid adaptability to diverse environmental stimuli. Such fast response times imply that adaptation cannot rely solely on synaptic plasticity, which operates on a much slower timescale. Instead, circuits must be inherently hyper-flexible and receptive to switches in functionalities. In this study, we show that a 4-neuron circuit can rapidly and controllably switch between 24 unique functions, while maintaining the same set of synaptic weights. Specifically, in order to systematically classify the outputs relative to inputs, we classify unique types of information processing in terms of 8 non-trivial logical truth tables (AND, OR, XOR, etc.). Furthermore, we test 3 different classes of input characteristics — difference in magnitude, timing and phase between input signals — and show that this small circuit can switch between different computations simply by adjusting its bias current. Finally, we demonstrate that this flexibility can be used to reduce the traditional nine gate adder into two 4-neuron circuits. This provides a computational foundation for how neural adaptability can occur on timescales much shorter than plasticity, an aspect important yet less explored in previous literature. This is a novel way to control a neural circuit, and could lead to new types of computing, especially in the growing field of neuromorphic computing.