# Madelon

[Browse datasets](<../../>) 

`libsvm:madelon@1` · Registry `1.4.0`. Default version in this registry. 

Artificial binary classification data from the NIPS 2003 feature selection challenge, distributed by LIBSVM with 500 features. The predefined test artifact contains the original challenge validation set, not its unlabeled test set. 

<a id="dimensions-and-tasks"></a>

## Dimensions and tasks

| Task | Type | Split | Samples | Predictors |
| --- | --- | --- | --- | --- |
| default | classification | test | 600 | 500 |
| default | classification | train | 2,000 | 500 |

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

<a id="retrieve"></a>

## 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. 

<a id="python"></a>

### Python

```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("libsvm:madelon@1", registry=registry)
```

<a id="r"></a>

### R

```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("libsvm:madelon@1", registry = registry)
```

<a id="julia"></a>

### Julia

```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("libsvm:madelon@1"; registry=registry)
```

<a id="cli"></a>

### CLI

```sh
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 'libsvm:madelon@1'
```

<a id="representation-and-variables"></a>

## Representation and variables

Tabular data with a sparse representation. Decoder: `libsvm-split` version 2. 

| Component | Kind | Type | Shape |
| --- | --- | --- | --- |
| train\_features | sparse\_matrix | float64 | 2,000 × 500 |
| train\_response | vector | int64 | 2,000 |
| test\_features | sparse\_matrix | float64 | 600 × 500 |
| test\_response | vector | int64 | 600 |

<a id="preprocessing"></a>

## Preprocessing

<a id="upstream"></a>

### Upstream

Status: documented. 

The exact numerical transformations used to prepare LIBSVM's feature values have not been established. 

**synthetic-generation.** Generate clusters on a five-dimensional hypercube, add 15 linear combinations to the five informative features and 480 distractor features, and randomize feature and observation order. 

**split-selection.** LIBSVM uses the original 600-row validation set as its test set because labels for the original 1800-row test set are unavailable. 

Evidence: [https://archive.ics.uci.edu/dataset/171/madelon](<https://archive.ics.uci.edu/dataset/171/madelon>), [https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/binary.html\#madelon](<https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/binary.html#madelon>) 

<a id="datamonger"></a>

### Datamonger

Status: none. 

LIBSVM v2 omits explicitly stored zeros without changing feature values. Datamonger applies no statistical preprocessing. 

Evidence: spec/decoders/libsvm-v2.md 

<a id="provenance-and-citation"></a>

## Provenance and citation

Provider: LIBSVM. Upstream name: madelon. Retrieved: 2026-10-26. 

[Original source](<https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/binary.html#madelon>) 

Authors: Isabelle Guyon. 

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

Original dataset and creator as cited by UCI. 

Isabelle Guyon, Steve Gunn, Asa Ben Hur, and Gideon Dror. Result analysis of the NIPS 2003 feature selection challenge. Advances in Neural Information Processing Systems, volume 17, 2005. 

Reference IG05a associated with Madelon by LIBSVM: [https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/ref.html\#IG05a](<https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/ref.html#IG05a>). 

<a id="license"></a>

## License

Unknown 

UCI licenses the original Madelon dataset under CC BY 4.0: [https://archive.ics.uci.edu/dataset/171/madelon](<https://archive.ics.uci.edu/dataset/171/madelon>). LIBSVM identifies that source and documents the split selection but states no license terms for its converted artifacts: [https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/binary.html\#madelon](<https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/binary.html#madelon>). The applicability of the original license to these distributed artifacts has not been established. 

<a id="artifacts"></a>

## Artifacts

Total download: 9.7 MiB. Sizes describe the distributed files, including any compression. 

| Artifact | Format | Compression | Size | Distribution | Preservation |
| --- | --- | --- | --- | --- | --- |
| train | libsvm | none | 7.4 MiB | upstream-only | unknown |
| test | libsvm | none | 2.2 MiB | upstream-only | unknown |

[train: upstream](<https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/binary/madelon>) 

[test: upstream](<https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/binary/madelon.t>) 

<a id="versions-in-this-registry"></a>

## Versions in this registry

[1](<1>) (default) 

<a id="registry-record"></a>

## Registry record

Download the [complete registry record](<../../../assets/576e90c2bb859a85e78cff86876c4d7eb446767acc68d2546733e77b9c9ef438.bin>), [registry selector](<../../../assets/34c66958a7fbdc23b69bacece50c1f1e9a097760400133967cecb6086963cdb1.bin>), or [full registry index](<https://github.com/jolars/datamonger/releases/download/registry-v1.4.0/index.json>). 

Index SHA-256: `ef83679266af9fcbb9ea194acdfb9bb6c380b440a574fc2ee7292899c5102909`.