Exoplanet Vit Temporal Eval

exoplanet-vit-temporal-eval

Exoplanet Classification through Vision Transformers with Temporal Image Analysis — Choudhary et al. (2025) (arXiv:2506.16597, 2025)

What this evaluates

Evaluates a Vision Transformer's ability to classify exoplanet transits by processing temporal light curve data transformed into image representations (Recurrence Plots and Gramian Angular Fields). It probes the model's capacity to capture long-range temporal dependencies and handle class imbalance in astronomical time-series data.

Datasets

• Kepler Light Curve Exoplanet Candidates — total ?; splits: 5-fold cross-validation (-1)

Metrics

• F1-score (primary) — range: [0, 1]
  • Harmonic mean of precision and recall: 2 * (precision * recall) / (precision + recall).
• Precision — range: [0, 1]
  • Ratio of true positive predictions to all positive predictions: TP / (TP + FP).
• Recall — range: [0, 1]
  • Ratio of true positive predictions to all actual positives: TP / (TP + FN).
• ROC value — range: [0, 1]
  • Area under the Receiver Operating Characteristic curve, plotting True Positive Rate against False Positive Rate at various classification thresholds.

Input / output format

Input: 2D image representations of 1D Kepler light curve time-series data, specifically Recurrence Plots (RPs) or Gramian Angular Fields (GAFs), fed into a Vision Transformer.

Output: Binary classification label: 'planet' (Class 1) or 'not-planet' (Class 0).

Scoring recipe

def compute_metrics(y_true, y_pred):
    tp = sum(1 for t, p in zip(y_true, y_pred) if t == 1 and p == 1)
    fp = sum(1 for t, p in zip(y_true, y_pred) if t == 0 and p == 1)
    fn = sum(1 for t, p in zip(y_true, y_pred) if t == 1 and p == 0)
    precision = tp / (tp + fp) if (tp + fp) > 0 else 0.0
    recall = tp / (tp + fn) if (tp + fn) > 0 else 0.0
    f1 = 2 * precision * recall / (precision + recall) if (precision + recall) > 0 else 0.0
    return precision, recall, f1

Common pitfalls

• Class imbalance heavily skews predictions toward the majority 'not-planet' class, requiring explicit under-sampling or careful metric interpretation.
• Performance varies significantly between input representations (RPs vs. GAFs) and is sensitive to the under-sampling strategy applied.
• The study uses local light curves for image generation, which may limit the ViT's ability to capture full transit dynamics compared to global curves.

Evidence (verbatim from paper)

The recall, precision, and F1-score values are 0.8946, 0.8509, and 0.8722, respectively. These values indicate a balanced trade-off between precision and recall, highlighting the model's proficiency in minimizing false positives while effectively identifying true positives.

Citation

@misc{choudhary2025exoplanet,
  title={Exoplanet Classification through Vision Transformers with Temporal Image Analysis},
  author={Choudhary et al. (2025)},
  year={2025},
  note={arXiv:2506.16597}
}

• arXiv: 2506.16597