Simulating Message Passing via Spiking Neural Networks Using Logical Gates

NeurIPS 2024 Workshop BDU Submission57 Authors

Keywords: Message Passing, Spiking Neural Networks

Abstract: It is hypothesized that the brain functions as a Bayesian inference engine, continuously updating its beliefs based on sensory input and prior knowledge. Message passing is an effective method for performing Bayesian inference within graphical models. In this paper, we propose that the XOR and the Equality factor nodes, which are important components in binary message passing, can be realized through a series of logical operations within a spiking neural network framework. Spiking neural networks simulate the behavior of neurons in a more biologically plausible manner. By constructing these factor nodes with a series of logical operations, we achieve the desired results using a minimal number of neurons and synaptic connections, potentially advancing the development of biological neuron-based computation. We validate our approach with two experiments, demonstrating the alignment between our proposed network and the sum-product message-passing algorithm.

Submission Number: 57