将脉冲神经网络与决策Transformer结合，实现低功耗序列控制。
Spiking Decision Transformers: Local Plasticity, Phase-Coding, and Dendritic Routing for Low-Power Sequence Control
- 在自注意力模块中嵌入脉冲神经元，用代理梯度端到端训练。
- 单次决策仅产生不到10个脉冲，推理能耗降低超四数量级。
- 适合部署在嵌入式和可穿戴设备上的实时低功耗控制任务。
基于Transformer架构的强化学习智能体在序列决策任务中表现优异，但其对密集矩阵运算的依赖使其难以适用于能量受限的边缘平台。脉冲神经网络具有超低功耗、事件驱动推理的优势，但此前尚无研究能将脉冲动力学与基于回报的序列建模无缝融合。本文提出脉冲决策Transformer（SNN-DT），在每个自注意力模块中引入漏电积分-放电神经元，通过代理梯度端到端训练，并结合生物启发的三因子可塑性、相位编码的脉冲位置编码以及轻量级树突路由模块。实验结果表明，SNN-DT在经典控制基准测试（CartPole-v1、MountainCar-v0、Acrobot-v1、Pendulum-v1）上表现达到或超过标准决策Transformer水平，且每次决策产生的脉冲数少于10个，这一指标暗示单次推理能耗降低超过四个数量级。该工作实现了序列建模与类脑能效的融合，为嵌入式与可穿戴设备上的实时低功耗控制提供了新路径。
原文摘要 · Abstract (English)
Reinforcement learning agents based on Transformer architectures have achieved impressive performance on sequential decision-making tasks, but their reliance on dense matrix operations makes them ill-suited for energy-constrained, edge-oriented platforms. Spiking neural networks promise ultra-low-power, event-driven inference, yet no prior work has seamlessly merged spiking dynamics with return-conditioned sequence modeling. We present the Spiking Decision Transformer (SNN-DT), which embeds Leaky Integrate-and-Fire neurons into each self-attention block, trains end-to-end via surrogate gradients, and incorporates biologically inspired three-factor plasticity, phase-shifted spike-based positional encodings, and a lightweight dendritic routing module. Our implementation matches or exceeds standard Decision Transformer performance on classic control benchmarks (CartPole-v1, MountainCar-v0, Acrobot-v1, Pendulum-v1) while emitting fewer than ten spikes per decision, an energy proxy suggesting over four orders-of-magnitude reduction in per inference energy. By marrying sequence modeling with neuromorphic efficiency, SNN-DT opens a pathway toward real-time, low-power control on embedded and wearable devices.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。