Solving constraint satisfaction problems with networks of spiking neurons

Abstract

Read online

Network of neurons in the brain apply – unlike processors in our current generation ofcomputer hardware – an event-based processing strategy, where short pulses (spikes) areemitted sparsely by neurons to signal the occurrence of an event at a particular point intime. Such spike-based computations promise to be substantially more power-efficient thantraditional clocked processing schemes. However it turned out to be surprisingly difficult todesign networks of spiking neurons that can solve difficult computational problems on the levelof single spikes (rather than rates of spikes). We present here a new method for designingnetworks of spiking neurons via an energy function. Furthermore we show how the energyfunction of a network of stochastically firing neurons can be shaped in a quite transparentmanner by composing the networks of simple stereotypical network motifs. We show that thisdesign approach enables networks of spiking neurons to produce approximate solutions todifficult (NP-hard) constraint satisfaction problems from the domains of planning/optimizationand verification/logical inference. The resulting networks employ noise as a computationalresource. Nevertheless the timing of spikes (rather than just spike rates) plays an essential rolein their computations. Furthermore, networks of spiking neurons carry out for the Traveling Salesman Problem a more efficient stochastic search for good solutions compared with stochastic artificial neural networks (Boltzmann machines) and Gibbs sampling.

Keywords

• neuromorphic hardware
• spiking neural networks
• neural sampling
• Boltzmann machine
• Benchmark tasks
• NP complete problems