Madelon: original UCI challenge splits

datamonger:guyon2004-madelon@1 · Registry 1.4.0. Default version in this registry.

Synthetic binary classification benchmark preserving all 500 features and the original training, validation, and unlabeled test splits.

Dimensions and tasks

Task	Type	Split	Samples	Predictors
default	classification	test	1,800	500
default	classification	train	2,000	500
default	classification	validation	600	500

Predictor counts follow each task's feature selection and exclude targets and identifiers. Sample counts are reported separately for each split.

Retrieve

These examples pin both the dataset version and the registry's exact index. An uncached dataset requires network access to its source. The CLI retrieves verified artifacts. The language clients also decode them.

Python

from datamonger import Registry, fetch_data

registry = Registry(
    release="1.4.0",
    index_sha256="ef83679266af9fcbb9ea194acdfb9bb6c380b440a574fc2ee7292899c5102909",
    index_url="https://github.com/jolars/datamonger/releases/download/registry-v1.4.0/index.json",
)
data = fetch_data("datamonger:guyon2004-madelon@1", registry=registry)

R

library(datamonger)

registry <- registry_selector(
  release = "1.4.0",
  index_sha256 = "ef83679266af9fcbb9ea194acdfb9bb6c380b440a574fc2ee7292899c5102909",
  index_url = "https://github.com/jolars/datamonger/releases/download/registry-v1.4.0/index.json"
)
data <- fetch_data("datamonger:guyon2004-madelon@1", registry = registry)

Julia

using Datamonger

registry = registry_selector(
    "1.4.0",
    "ef83679266af9fcbb9ea194acdfb9bb6c380b440a574fc2ee7292899c5102909",
    "https://github.com/jolars/datamonger/releases/download/registry-v1.4.0/index.json",
)
data = fetch_data("datamonger:guyon2004-madelon@1"; registry=registry)

CLI

cat > datamonger-registry.json <<'JSON'
{
  "index_sha256": "ef83679266af9fcbb9ea194acdfb9bb6c380b440a574fc2ee7292899c5102909",
  "index_url": "https://github.com/jolars/datamonger/releases/download/registry-v1.4.0/index.json",
  "release": "1.4.0",
  "schema_version": 1
}
JSON
datamonger --selector datamonger-registry.json fetch 'datamonger:guyon2004-madelon@1'

Representation and variables

Tabular data with a dense representation. Decoder: named-splits version 1.

Component	Kind	Type	Shape
train/x1	vector	int64	2,000
train/x2	vector	int64	2,000
train/x3	vector	int64	2,000
train/x4	vector	int64	2,000
train/x5	vector	int64	2,000
train/x6	vector	int64	2,000
train/x7	vector	int64	2,000
train/x8	vector	int64	2,000
train/x9	vector	int64	2,000
train/x10	vector	int64	2,000
train/x11	vector	int64	2,000
train/x12	vector	int64	2,000
train/x13	vector	int64	2,000
train/x14	vector	int64	2,000
train/x15	vector	int64	2,000
train/x16	vector	int64	2,000
train/x17	vector	int64	2,000
train/x18	vector	int64	2,000
train/x19	vector	int64	2,000
train/x20	vector	int64	2,000
train/x21	vector	int64	2,000
train/x22	vector	int64	2,000
train/x23	vector	int64	2,000
train/x24	vector	int64	2,000
train/x25	vector	int64	2,000
train/x26	vector	int64	2,000
train/x27	vector	int64	2,000
train/x28	vector	int64	2,000
train/x29	vector	int64	2,000
train/x30	vector	int64	2,000

Showing 30 of 1,502 components. The complete registry record contains every variable name and task binding.

Preprocessing

Upstream

Status: documented.

synthetic-generation. Generate five informative dimensions, 15 redundant linear combinations, and 480 distractors; randomize feature and observation order.

Evidence: https://archive.ics.uci.edu/dataset/171/madelon

Datamonger

Status: none.

Lossless CSV conversion and gzip compression only. Preserve all values, labels, row and feature order, and splits.

Evidence: curation/guyon2004-madelon/1/prepare.py

Provenance and citation

Provider: Datamonger; derived from UCI. Upstream name: Madelon. Retrieved: 2026-10-26.

Original source

Authors: Isabelle Guyon.

Guyon, I. (2004). Madelon [Dataset]. UCI Machine Learning Repository. https://doi.org/10.24432/C5602H.

Madelon: original UCI challenge splits (Datamonger representation, version 1). https://doi.org/10.5281/zenodo.23665387.

License

CC-BY-4.0

License information

UCI explicitly licenses Madelon under CC BY 4.0: https://archive.ics.uci.edu/dataset/171/madelon.

Artifacts

Total download: 2.6 MiB. Sizes describe the distributed files, including any compression.

Artifact	Format	Compression	Size	Distribution	Preservation
train	csv	gzip	1.2 MiB	mirror	durable
validation	csv	gzip	365.9 KiB	mirror	durable
test	csv	gzip	1.1 MiB	mirror	durable

Versions in this registry

1 (default)

Registry record

Download the complete registry record, registry selector, or full registry index.

Index SHA-256: ef83679266af9fcbb9ea194acdfb9bb6c380b440a574fc2ee7292899c5102909.