This document outlines the planned development milestones for SpikingTSF.
- Unified training and evaluation framework
- 11 SNN models + 2 ANN baselines integrated
- ETTh1 and ETTh2 datasets supported
- Hyperparameter tuning infrastructure (Optuna)
- Multi-seed evaluation with mean ± std reporting
- Output logging (per-run + aggregated)
- Documentation files (README, CONTRIBUTING, PROTOCOL, DATASETS, REPRODUCIBILITY, MODEL_ZOO, etc.)
- CITATION.cff and citation policy
- GitHub repository transferred to Spikora organization
- CI workflow for import checks
-
tests/directory with basic sanity checks - Verified leaderboard results for all 13 models on ETTh1 and ETTh2
- Full RESULTS.md with results for all models at all horizons (96 / 192 / 336 / 720)
- Hyperparameter tuning complete for ETTm1 and ETTm2
- Full results for all 4 ETT datasets and all 4 horizons
- Updated leaderboard in README
- Weather dataset support
- Electricity dataset support
- Traffic and/or PEMS datasets (if within scope)
- Results on additional datasets
- Efficiency metrics: spike rate, synaptic operations (SOPs), parameter count, and theoretical energy estimation.
- Potential support for neuromorphic hardware evaluation if infrastructure becomes available.
- Additional SNN architectures from newer publications.
Items without checkboxes are planned but not yet in progress. The roadmap may shift as research priorities evolve.