Research map / Architectures and efficiency / Spiking networks
Reconstructing Spiking Neural Networks Using a Single Neuron with Autapses
Wuque Cai, Hongze Sun, Quan Tang, Shifeng Mao, Zhenxing Wang, Jiayi He, Duo Chen, Dezhong Yao, Daqing Guo
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace standard SNN reservoir with TDA-SNN at small size (16 nodes) (ablation) | TDA-SNN · DEAP | accuracy (%): 81.21 -> 77.59 | Hurt |
| scale reservoir to 256 nodes with TDA-SNN vs STD-SNN (ablation) | TDA-SNN · DEAP | accuracy (%): 79.92 -> 88.65 | Improved |
| scale reservoir to 256 nodes with TDA-SNN vs STD-SNN (ablation) | TDA-SNN · SHD | accuracy (%): 77.63 -> 80.04 | Improved |
| increase SHD reservoir from 128 to 256 nodes (ablation) | TDA-SNN · SHD | accuracy (%): 80.68 -> 80.04 | Hurt |
| max-connection delay selection at 2 delays vs random delay (ablation) | TDA-SNN RC · SHD | accuracy (%): 78.62 -> 72.09 | Hurt |
| increase number of autapse delays 1 to 64 (random delay) (ablation) | TDA-SNN MLP · MNIST | accuracy (%): 84.03 -> 96.42 | Improved |
| random delay vs max-connection selection with many delays (ablation) | TDA-SNN MLP · MNIST | accuracy (%): 94.38 -> 96.42 | Improved |
| random delay vs max-connection selection with many delays (ablation) | TDA-SNN RC · DEAP | accuracy (%): 84.47 vs 84.47 at 64 delays | No clear effect |
| increase MLP depth 1 to 4 layers (ablation) | TDA-SNN MLP · MNIST | accuracy (%): 94.41 -> 95.60 | Improved |
| increase MLP depth 1 to 4 layers (ablation) | TDA-SNN MLP · fMNIST | accuracy (%): 86.62 -> 87.12 then decreases | Mixed |
| increase MLP nodes from 16 to 256 (ablation) | TDA-SNN MLP · DVS Gesture | accuracy (%): 53.99 -> 72.95 | Improved |
| use TDA-SNN single-neuron convolution instead of standard SNN conv | TDA-SNN · CIFAR-10 | accuracy (%): 47.31 -> 37.64 | Hurt |
Closest papers by meaning
- High-Performance Temporal Reversible Spiking Neural Networks with $\mathcal{O}(L)$ Training Memory and $\mathcal{O}(1)$ Inference Cost (ICML 2024 · similarity 0.89)
- Enhanced Self-Distillation Framework for Efficient Spiking Neural Network Training (NeurIPS 2025 · similarity 0.89)
- Spike-driven Transformer V2: Meta Spiking Neural Network Architecture Inspiring the Design of Next-generation Neuromorphic Chips (ICLR 2024 · similarity 0.89)
- Memory-Efficient Reversible Spiking Neural Networks (AAAI 2024 · similarity 0.89)
- Multiplication-Free Parallelizable Spiking Neurons with Efficient Spatio-Temporal Dynamics (NeurIPS 2025 · similarity 0.89)
- SpikingSSMs: Learning Long Sequences with Sparse and Parallel Spiking State Space Models (AAAI 2025 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
