Network of evolvable neural units can learn synaptic learning rules and spiking dynamics

Abstract

Although deep neural networks have seen great success in recent years through various changes in overall architectures and optimization strategies, their fundamental underlying design remains largely unchanged. Computational neuroscience may provide more biologically realistic models of neural processing mechanisms, but they are still high-level abstractions of empirical behaviour. Here we propose an evolvable neural unit (ENU) that can evolve individual somatic and synaptic compartment models of neurons in a scalable manner. We demonstrate that ENUs can evolve to mimic integrate-and-fire neurons and synaptic spike-timing-dependent plasticity. Furthermore, by constructing a network where an ENU takes the place of each synapse and neuron, we evolve an agent capable of learning to solve a T-maze environment task. This network independently discovers spiking dynamics and reinforcement-type learning rules, opening up a new path towards biologically inspired artificial intelligence. Bertens and Lee propose an evolvable neural unit, a recurrent neural network-based module that can evolve individual somatic and synaptic compartment models of neurons. By constructing networks of these evolvable neural units, they can evolve agents that learn synaptic update rules and the spiking dynamics of neurons.