Gw Glitch Mitigation Eval

---
name: gw-glitch-mitigation-eval
description: Evaluates the ability of a joint signal-glitch model to accurately recover compact binary coalescence parameters and reconstruct glitch waveforms when they overlap in real gravitational-wave detector data. It compares a baseline model ignoring glitches against a full model that jointly fits both components. Use when the user wants to benchmark on LIGO O3 data segments, or asks about evaluating this task. Reports mismatch.
metadata:
  skill_kind: dataset_eval
  source_arxiv: 2205.13580
  bibtex_key: hourihane2022gwglitch
  confidence: high
---

# gw-glitch-mitigation-eval

> Accurate modeling and mitigation of overlapping signals and glitches in gravitational-wave data — Hourihane et al. (2022) (arXiv:2205.13580, 2022)

## What this evaluates

Evaluates the ability of a joint signal-glitch model to accurately recover compact binary coalescence parameters and reconstruct glitch waveforms when they overlap in real gravitational-wave detector data. It compares a baseline model ignoring glitches against a full model that jointly fits both components.

## Datasets

- **LIGO O3 data segments** — total ?; splits: test (-1)

## Metrics

- `mismatch` **(primary)** — range: [0, 1]
  - 1 minus the maximum overlap between the recovered and injected waveforms optimized over time and phase shifts. Lower values indicate more faithful waveform recovery.
- `overlap` — range: [0, 1]
  - Maximum inner product between two waveforms normalized by their norms, optimized over time and phase shifts. Measures morphological similarity between signal and glitch components.

## Input / output format

**Input**: Real LIGO detector strain data segments (1–2 minutes) containing injected compact binary coalescence signals overlapping with instrumental glitches (blip, slow-scattering, or fast-scattering).

**Output**: Marginalized posterior distributions for detector-frame chirp mass, luminosity distance, and effective spin, plus reconstructed glitch waveforms via Bayesian sampling.

## Scoring recipe

```python
def compute_mismatch(rec, gold):
    ip = inner_product(rec, gold)
    norm = sqrt(inner_product(rec, rec) * inner_product(gold, gold))
    return 1 - max_over_time_phase(ip / norm)

def compute_overlap(rec, gold):
    ip = inner_product(rec, gold)
    norm = sqrt(inner_product(rec, rec) * inner_product(gold, gold))
    return max_over_time_phase(abs(ip / norm))
```

## Common pitfalls

- Mismatch and overlap are evaluated on fixed injected parameters rather than marginalized over posteriors, making them conservative estimates of recovery quality.
- Glitch reconstruction quality depends heavily on SNR and prior tuning (e.g., wavelet amplitude peak), requiring case-by-case attention for low-SNR glitches.

## Evidence (verbatim from paper)

> The set in the first column measure how well the models were reconstructed: the lower the mismatch, the more faithful each model recovery is. Specifically in all cases we find ${\mathcal{M}}(g_{\textrm{G}}|g_{\textrm{CBC+G}})\leq 0.01$, suggesting that the recovered glitch model does not consume any significant amount of the injected CBC signal, nor does it miss part of the glitch due to the presence of the signal. All overlaps are small, however these is some clear variation. Specifically, we find ${\mathcal{O}}(g_{\textrm{G}}|h^{\textrm{inj}})\sim{\mathcal{O}}(g_{\textrm{G}}|h_{\textrm{CBC+G}}^{\textrm{rec}})\leq{\mathcal{O}}(h_{\textrm{CBC}}^{\textrm{rec}}|g_{\textrm{G}})$ which means that in the “CBCOnly” analysis the CBC model absorbs part of the glitch power.

## Citation

```bibtex
@misc{hourihane2022gwglitch,
  title={Accurate modeling and mitigation of overlapping signals and glitches in gravitational-wave data},
  author={Hourihane et al. (2022)},
  year={2022},
  note={arXiv:2205.13580}
}
```

- arXiv: 2205.13580