Summary
Learning in the brain happens through tiny electrical signals called spikes, but how brain circuits assign credit for learning over time using only local spike information has remained unclear. The authors propose a new way to understand this learning problem as separating important signals from past events in the current brain state. They develop an algorithm called gradient tunneling that learns from the timing of spikes and works well with both brain-like and artificial neural networks. Their approach shows strong results on complex tasks involving memory and noisy information, while using fewer resources than existing methods. This work offers a new, plausible way to explain how the brain might solve the challenge of learning across time.
spiking neural networkstemporal credit assignmentneural microcircuitsspike timinggradient tunnelingstate separationevidence integrationnoise-robust memorybackpropagationlead-lag expansion
Authors
Xiangnan Zhang, Jingxin Liu, Ranqi Lu, Jingyu Liu, Qunxi Dong, Fuze Tian, Lixian Zhu, Bin Hu, Björn W. Schuller
Abstract
The brain uses discrete spikes for dynamic computation, yet, how neural microcircuits (NMCs) solve temporal credit assignment using local spike timing remains a fundamental open question. Dominant spiking neural network (SNN) approaches circumvent this by approximating backpropagation through surrogate gradients, decoupling learning from biological spike timing. Here, we reformulate temporal credit assignment as a state separation problem: extracting task-required components induced by historical perturbations directly from the current neural state. This enables an online feedback learning framework for NMCs through a gradient tunneling (GT) algorithm and the lead-lag expansion technique that derives credit assignment from local synaptic spike timing, while remaining compatible with ANN-SNN hybrid architectures. Experimentally, GT-trained NMCs excel at long-timescale evidence integration and noise-robust memory retention, and perform comparably to leading SNN online learning methods on real-world benchmarks with far fewer parameters. The proposed framework addresses the two-decade-old NMC feedback learning problem and suggests a computationally plausible explanation for the brain's learning mechanisms.