Research map / Architectures and efficiency / Spiking networks
Spike-inspired rank coding for fast and accurate recurrent neural networks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Train with rank coding and stop at the first threshold crossing | LSTM on the 2-sequence problem · 2-sequence problem | mean first-spike time: 23 -> 14 time-steps | Helped |
| Train with rank coding and stop at the first threshold crossing | LSTM on the 2-sequence problem · 2-sequence problem | accuracy: 96.84% -> 95.45% | Hurt |
| Use RC-training rather than conventional training | LSTM on temporally-coded MNIST · Temporally-coded MNIST, early inference | inference time | Helped |
| Add entropy-rewarding regularization with non-zero beta | RC-trained LSTM on Google Speech Commands · Google Speech Commands validation set | accuracy-speed trade-off: around 600 ms | Helped |
| Use beta=0 instead of a non-zero entropy regularizer (ablation) | RC-trained LSTM on Google Speech Commands · Google Speech Commands validation set | inference time: less than 500 ms; 1000 ms | Helped |
| Use beta=0 instead of non-zero entropy regularization (ablation) | RC-trained LSTM on Google Speech Commands · Google Speech Commands validation set | accuracy | Hurt |
| Apply beta regularization without RC-training (ablation) | Conventionally trained LSTM on Google Speech Commands · Google Speech Commands validation set | speed-accuracy trade-off | Hurt |
| Vary the threshold used during RC-training (ablation) | RC-trained LSTM on the 2-sequence problem · 2-sequence problem | speed-accuracy trade-off | No clear effect |
| Use RC-training rather than conventional training | LSTM on continuous sequence spotting · Continuous sequence spotting | accuracy: 100% | Helped |
| Use RC-training rather than conventional training | LIF spiking network on temporal MNIST · Temporal MNIST | early-spike accuracy | Helped |
Closest papers by meaning
- Temporal Weighted Encoding: Towards Maximal-Capacity Spike Coding for ANN–SNN Conversion (ICML 2026 · similarity 0.91)
- Advancing Training Efficiency of Deep Spiking Neural Networks through Rate-based Backpropagation (NeurIPS 2024 · similarity 0.90)
- Temporal Flexibility in Spiking Neural Networks: Towards Generalization Across Time Steps and Deployment Friendliness (ICLR 2025 · similarity 0.89)
- SpikingSSMs: Learning Long Sequences with Sparse and Parallel Spiking State Space Models (AAAI 2025 · similarity 0.89)
- Learning Delays in Spiking Neural Networks using Dilated Convolutions with Learnable Spacings (ICLR 2024 · similarity 0.88)
- Gated Attention Coding for Training High-Performance and Efficient Spiking Neural Networks (AAAI 2024 · similarity 0.88)
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.