Reproduce “Proxy Target: Bridging the Gap Between Discrete Spiking Neural Networks and Continuous Control” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Zijie Xu, Tong Bu, Zecheng Hao, Jianhao Ding, Zhaofei Yu
The authors introduce a proxy-target method for training spiking neural network actors in continuous-control reinforcement learning. A continuous proxy imitates the online actor and supplies smoother target actions, while the deployed actor remains spiking. Across five MuJoCo tasks, proxy-target LIF gains 32.15% average performance over vanilla LIF, but performs worse on Hopper; its aggregate gain over ANN TD3 is 5.84%.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.