# Exoplanet Anomaly Detection Eval

> Evaluates the ability of autoencoder-derived latent representations combined with various anomaly detection algorithms to identify chemically anomalous exoplanet transit spectra under realistic observational noise levels. It benchmarks reconstruction loss, 1-class SVM, K-means, and LOF across raw spectral and latent feature spaces. Use when the user wants to benchmark on Exoplanet transit spectra database, or asks about evaluating this task. Reports AUC.

- Skill: `qhjqhj00/exoplanet-anomaly-detection-eval` (Agent Skill)
- Install (CLI): `npx skillmds add qhjqhj00/exoplanet-anomaly-detection-eval`
- Raw SKILL.md: https://api.skillmd.com/api/skills/qhjqhj00/exoplanet-anomaly-detection-eval/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Productivity
- Author: qhjqhj00 (https://skillmd.com/u/qhjqhj00)
- Updated: 2026-09-08
- Page: https://skillmd.com/skills/qhjqhj00/exoplanet-anomaly-detection-eval

---


# exoplanet-anomaly-detection-eval

> Hunting for "Oddballs" with Machine Learning: Detecting Anomalous Exoplanets Using a Deep-Learned Low-Dimensional Representation of Transit Spectra with Autoencoders — Roman et al. (2026) (arXiv:2601.02324, 2026)

## What this evaluates

Evaluates the ability of autoencoder-derived latent representations combined with various anomaly detection algorithms to identify chemically anomalous exoplanet transit spectra under realistic observational noise levels. It benchmarks reconstruction loss, 1-class SVM, K-means, and LOF across raw spectral and latent feature spaces.

## Datasets

- **Exoplanet transit spectra database** — total ?; splits: test (-1)

## Metrics

- `AUC` **(primary)** — range: [0, 1]
  - Area under the Receiver Operating Characteristic (ROC) curve, which plots the true positive rate against the false positive rate across all classification thresholds.

## Input / output format

**Input**: High-resolution exoplanet transmission spectra with added Gaussian noise (10, 20, 30, or 50 ppm) and binary ground-truth labels (normal vs. anomalous).

**Output**: Anomaly scores or binary classification predictions (normal/anomalous) for each spectrum.

## Scoring recipe

```python
def compute_auc(y_true, y_scores):
    fpr, tpr, _ = roc_curve(y_true, y_scores)
    return auc(fpr, tpr)
```

## Common pitfalls

- Performance degrades significantly as noise increases beyond 30 ppm, especially when methods are applied directly to raw spectral features instead of latent representations.
- Comparing anomaly detection algorithms across different feature spaces (spectral vs. latent) without controlling for noise levels can lead to misleading conclusions about algorithm superiority.

## Evidence (verbatim from paper)

> The anomaly detection performance was studied in two different spaces: the spectral space and the latent space. We found that anomaly detection works better in the latent space. All models presented in this study show higher AUC values in latent space compared to spectral space.

## Citation

```bibtex
@misc{roman2026oddballs,
  title={Hunting for "Oddballs" with Machine Learning: Detecting Anomalous Exoplanets Using a Deep-Learned Low-Dimensional Representation of Transit Spectra with Autoencoders},
  author={Roman et al. (2026)},
  year={2026},
  note={arXiv:2601.02324}
}
```

- arXiv: 2601.02324