fuzzy2snn

Reproducible benchmark suite for neuro-fuzzy controllers with a spiking Sugeno backend (publish-mode: CSV/log/manifest/Pareto plot).

What’s inside

• FIS zoo fetcher (MIT-licensed controllers where possible)
• Fuzzy → SNN compilation + spiking Sugeno backend
• Publish-mode evaluation pipeline:
  • clean CSV output
  • separate log output
  • JSON manifest
  • Pareto plot

Quick start

pip install -r requirements.txt
python examples/fis_zoo_benchmark.py --help

Reproducibility (Paper Mode)

This repository supports a strict publish-mode designed for academic reproducibility.

The following command reproduces the results reported in the paper:

python examples/fis_zoo_benchmark.py \
  --controllers nickgkan_neurofuzzy3,nickgkan_neurofuzzy5 \
  --samples 200 \
  --sweep_sugeno_spiking \
  --sweep_dt_ms_list 0.1,0.2 \
  --sweep_control_window_ms_list 100 \
  --sweep_rule_max_rate_hz_list 470,1000,2000 \
  --mae_targets 0.25,0.20,0.10 \
  --tune_seeds 1,2,3,4,5 \
  --eval_seeds 11,12,13,14,15,16,17,18,19,20 \
  --replay_best \
  --csv_path out.csv \
  --log_path out.log \
  --save_manifest run_manifest.json \
  --plot_pareto_path pareto.png