Supervised Spike Agreement Dependent Plasticity for Fast Local Learning in Spiking Neural Networks

TL;DR

The work addresses the challenge of supervised learning in spiking neural networks (SNNs) without backpropagation by introducing Supervised Spike Agreement--Dependent Plasticity (SADP), which uses population-level agreement metrics like Cohen’s $\kappa$ to drive local, linear-time updates. By coupling a frozen CNN front-end with Poisson-encoded spike trains, the approach scales from grayscale MNIST-like data to color and biomedical images while preserving locality in synaptic updates. Across standard vision benchmarks and biomedical datasets, the method achieves competitive accuracy, fast convergence, and robust performance under limited hyperparameter tuning, with demonstrated hardware compatibility via device-inspired update kernels. The results advocate a principled, hardware-aligned pathway for scalable, interpretable spike-based supervised learning that complements gradient-based methods rather than replacing them, and they point to future work on deeper architectures and temporally structured tasks.

Abstract

Spike-Timing-Dependent Plasticity (STDP) provides a biologically grounded learning rule for spiking neural networks (SNNs), but its reliance on precise spike timing and pairwise updates limits fast learning of weights. We introduce a supervised extension of Spike Agreement-Dependent Plasticity (SADP), which replaces pairwise spike-timing comparisons with population-level agreement metrics such as Cohen's kappa. The proposed learning rule preserves strict synaptic locality, admits linear-time complexity, and enables efficient supervised learning without backpropagation, surrogate gradients, or teacher forcing. We integrate supervised SADP within hybrid CNN-SNN architectures, where convolutional encoders provide compact feature representations that are converted into Poisson spike trains for agreement-driven learning in the SNN. Extensive experiments on MNIST, Fashion-MNIST, CIFAR-10, and biomedical image classification tasks demonstrate competitive performance and fast convergence. Additional analyses show stable performance across broad hyperparameter ranges and compatibility with device-inspired synaptic update dynamics. Together, these results establish supervised SADP as a scalable, biologically grounded, and hardware-aligned learning paradigm for spiking neural networks.

Figures (5)

• Figure 1: Pastel-colored experimental workflow: Poisson or CNN+Poisson encoding feeds into evaluation, which then connects to supervised SADP learning.
• Figure 3: Optimization history plot showing the evolution of validation accuracy across 50 Optuna trials. Blue dots indicate individual trial accuracies, while the orange line represents the cumulative best value. Significant improvements occur around Trials 4, 12–14, and 46–48, reaching a best accuracy of approximately 0.57.
• Figure 4: Hyperparameter importance plot illustrating the relative contribution of each parameter to model performance. Learning rate decay dominates with an importance of 0.62, followed by output learning rate ($\eta_{out}$) and hidden layer size ($N_{hid}$).
• Figure 5: Device-inspired SADP.
• Figure : (a) FMNIST