Research map / Time series / Long-term forecasting
A Spiking Heterogeneous Harmonic Resonate-and-Fire State Space Model for Time Series
Kartikay Agrawal, Vaishnavi Nagabhushana, Abhijeet Vikram, Vedant Sharma, Ayon Borthakur
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 |
|---|---|---|---|
| add heterogeneous learnable parameters to SH2RFSSM (ablation) | SH2RFSSM · UEA six-dataset average | accuracy (%): 61.7 -> 67.5 | Improved |
| make all parameters heterogeneous instead of a subset (ablation) | SH2RFSSM · UEA six-dataset average | accuracy (%): 67.5 -> 65.9 | Hurt |
| add learnable damping b (ablation) | SH2RFSSM · UEA six-dataset average | accuracy (%): 64.2 -> 63.5 | Hurt |
| use IM discretisation instead of explicit with heterogeneity in BHRF (ablation) | BHRF · SHD | SOPs: 4139.5 -> 2937.7 | Improved |
| use IM or IMEX discretisation instead of EX in HRF networks (ablation) | BHRF · SHD | accuracy (%): 93.6±0.3 -> 93.5±0.7 | No clear effect |
| use IM discretisation instead of IMEX (ablation) | SH2RFSSM · UEA six-dataset average | accuracy (%): 65.3 -> 66.5 | Improved |
| use IMEX instead of IM discretisation (ablation) | SH2RFSSM · PPG-DaLiA | MSE x10^-2: 11.8 -> 8.5 | Improved |
| add learnable temporal-kernel spiking regressor with HRF layers | SH2RFSSM · PPG-DaLiA | MSE: Mamba 10.7 -> 8.5 | Improved |
| replace FFT/IFFT with HRF layer | SpikF · ETTh1 efficiency | latency and energy: 7.77 ms -> 3.93 ms; 7.84 µJ -> 1.98 µJ | Improved |
| replace FFT/IFFT with HRF layer | SpikF · ETTh1 and ETTm1 | MSE: 0.44 -> 0.443 | No clear effect |
| replace FFT/IFFT with HRF layer | SpikF · ECL | MSE: 0.183 -> 0.195 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- iTransformer — iTransformer: Inverted Transformers Are Effective for Time Series Forecasting (ICLR 2024) · better on 0 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- SpikingSSMs: Learning Long Sequences with Sparse and Parallel Spiking State Space Models (AAAI 2025 · similarity 0.90)
- Balanced Resonate-and-Fire Neurons (ICML 2024 · similarity 0.89)
- LongSpike: Fractional Order Spiking State Space Models for Efficient Long Sequence Learning (NeurIPS 2026 · similarity 0.88)
- SpikF: Spiking Fourier Network for Efficient Long-term Prediction (ICML 2025 · similarity 0.87)
- Dendritic Resonate-and-Fire Neuron for Effective and Efficient Long Sequence Modeling (NeurIPS 2025 · similarity 0.87)
- P-SPIKESSM: HARNESSING PROBABILISTIC SPIKING STATE SPACE MODELS FOR LONG-RANGE DEPENDENCY TASKS (ICLR 2025 · similarity 0.87)
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.
