This repository contains a fresh ICML-style workshop draft for the Meta-MAPG two-phase convergence and restart-globalisation thesis.
Canonical files:
main.tex: LaTeX source for the paper.main.pdf: compiled PDF.references.bib: bibliography.experiments/run_meta_mapg_experiments.py: sample-based tabular experiments.experiments/run_mlp_ipd.py: optional PyTorch MLP-policy IPD check.artifacts/main/: experiment outputs used in the paper.artifacts/mlp/: aggregate MLP-policy IPD check outputs.figures/andtables/: paper-ready copies of the generated figures and summary table.
Regenerate the experiments from this directory with:
python3 experiments/run_meta_mapg_experiments.py \
--outdir artifacts/main \
--seeds 100 \
--steps 260 \
--restart-steps 120 \
--max-restarts 12 \
--selection-budget 12 \
--selection-seeds 100 \
--selection-steps 120 \
--trajectory-steps 140 \
--trajectory-batch-size 384 \
--trajectory-grid-size 5 \
--batch-size 384 \
--basin-batch-size 192 \
--grid-size 21 \
--basin-steps 140 \
--reference-batch-size 120000 \
--sanity-reps 80 \
--own-coef 0.35 \
--peer-coef 1.5Regenerate the optional MLP-policy IPD appendix check with PyTorch installed:
python3 experiments/run_mlp_ipd.py \
--n_seeds 100 \
--n_steps 260 \
--batch_size 384 \
--peer_coef 1.5 \
--own_coef 0.35 \
--inner_lr 0.55 \
--out_dir artifacts/mlpRegenerate the appendix constant-vs-two-phase MLP comparison with:
python3 experiments/run_mlp_ipd.py \
--skip_ablation \
--run_annealing_compare \
--anneal_seeds 20 \
--anneal_steps 2000 \
--anneal_phase1_steps 100 \
--anneal_scale 30 \
--anneal_power 0.7 \
--anneal_log_every 20 \
--batch_size 384 \
--peer_coef 1.5 \
--own_coef 0.35 \
--inner_lr 0.55 \
--out_dir artifacts/mlpRecompile the paper with:
latexmk -pdf -interaction=nonstopmode main.tex