Noise Diversification

Noise family selection, mixture modes, and per-dataset resolution.

Use noise-family workflows when you want non-Gaussian stochastic regimes while retaining deterministic seed behavior and explicit metadata reporting.

When to use

  • You want heavier-tail stochasticity than the Gaussian default.
  • You need deterministic comparisons across Gaussian/Laplace/Student-t regimes.
  • You want benchmark guardrails for runtime impact and metadata validity.

Supported families

  • gaussian: default Gaussian sampling.
  • laplace: heavier-tailed Laplace noise.
  • student_t: heavy-tailed Student-t (df > 2).
  • mixture: weighted mixture over Gaussian/Laplace/Student-t.

Preset workflows

Generate smoke datasets for each family:

dagzoo generate --config configs/preset_noise_gaussian_generate_smoke.yaml --num-datasets 25 --out data/run_noise_gaussian
dagzoo generate --config configs/preset_noise_laplace_generate_smoke.yaml --num-datasets 25 --out data/run_noise_laplace
dagzoo generate --config configs/preset_noise_student_t_generate_smoke.yaml --num-datasets 25 --out data/run_noise_student_t
dagzoo generate --config configs/preset_noise_mixture_generate_smoke.yaml --num-datasets 25 --out data/run_noise_mixture

Benchmark guardrail smoke run:

dagzoo benchmark \
  --config configs/preset_noise_benchmark_smoke.yaml \
  --preset custom \
  --suite smoke \
  --no-memory \
  --out-dir benchmarks/results/smoke_noise

What to inspect

  • Per-dataset metadata.ndjson entries:
    • noise_distribution.family_requested
    • noise_distribution.family_sampled
    • noise_distribution.sampling_strategy
    • noise_distribution.base_scale
    • noise_distribution.student_t_df
    • noise_distribution.mixture_weights (when requested family is mixture)
  • Benchmark summary noise_guardrails:
    • metadata coverage/validity
    • sampled-family counts
    • runtime delta vs gaussian-noise control

For output details, see output-format.md.