Skip to main content

Datasets

Status:ValidIn progress
Tags:Dataset Registry
Last modified:

This section is the registry of datasets used by πŸš‚ENABOL ablation experiments. Each dataset gets a stable ID so experiment pages can reference the input distribution by name instead of re-explaining the full construction.

Use dataset IDs in experiment trace blocks:

Dataset: DS-AFFINE-LINEAR-000

Dataset Table​

Dataset IDNameTypeInputStatusUsed by
DS-AFFINE-LINEAR-000Affine linear, no biasSynthetic regressionx∼U([0,1]4)x \sim \mathcal{U}([0,1]^4)ValidEXP-000A and EXP-000B.

ID Convention​

Use:

DS-<FAMILY>-<TASK>-<NUMBER>

Examples:

DS-AFFINE-LINEAR-000
DS-AFFINE-BIAS-001
DS-SENSOR-DRIFT-010

Keep IDs stable once they are used in an experiment page. If the data generation changes in a way that affects interpretation, create a new dataset ID.

Dataset Record Checklist​

Each dataset page should define:

  • ID and short name,
  • generation equations,
  • dimensions,
  • train/test or online sampling procedure,
  • drift/corruption model,
  • default random seed if relevant,
  • links to plots or notebooks,
  • caveats that affect interpretation.