bilby-laplace 0.2.0__tar.gz

bilby_laplace-0.2.0/.gitignore

@@ -0,0 +1,51 @@
+# setuptools-scm auto-generated; do not commit
+src/bilby_laplace/_version.py
+
+# Build / packaging
+dist/
+build/
+*.egg-info/
+**/__pycache__/
+*.py[cod]
+*.so
+
+# Common runtime outputs
+outdir/
+*.png
+*.npz
+
+# macOS
+.DS_Store
+
+# Vim
+*.swp
+*.swo
+
+# Examples
+*outdir*
+
+# Paper working directory (studies, runs, figures) - kept out of the package repo
+paper/
+
+# MkDocs build output
+site/
+
+# Test / coverage artifacts
+.coverage
+.coverage.*
+coverage.xml
+htmlcov/
+.pytest_cache/
+
+# Claude Code guidance - local only, kept out of git history
+CLAUDE.md
+.claude/
+
+# Run logs from `make <sampler> > x.log`
+examples/**/*.log
+
+# Editor settings
+.vscode/
+
+# Example outputs: `make compare` writes a results README and figures here
+examples/*/README.md

bilby_laplace-0.2.0/CHANGELOG.md

@@ -0,0 +1,112 @@
+# Changelog
+
+All notable changes to this project will be documented in this file.
+
+The format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
+Versions correspond to git tags; version numbers follow
+[Semantic Versioning](https://semver.org/).
+
+---
+
+## [Unreleased]
+
+## [0.2.0] — 2026-11-11
+
+### Added
+
+- Resampling methods: `resample="smc"` (via [aspire](https://github.com/bilby-dev/aspire)),
+  `"inprior"` (draw from the proposal within the prior support, no reweighting) and `"emcee"`.
+- Multimodal proposals: `n_modes`, with `mode_searches` (`"hypercube"`, `"symmetric"`,
+  `"symmetric-polish"`, `"multistart"`), `mode_search_nsamples`, `mode_search_subspace`,
+  `mode_separation_sigma`, `mode_symmetries`, `mode_symmetry_tol`, `mode_symmetry_polish_max`,
+  `mode_multistart_nstarts` and `mode_weights`. Candidates from all searches are pooled and ranked
+  by log-posterior before selection. The mixture is recorded in `meta_data["mode_mixture"]`,
+  including which search proposed each mode.
+- `proposal_covariance`: `"diagonal"` (independent truncated normals) or `"full"` (the correlated
+  Gaussian restricted to the prior box). Recorded in `meta_data["proposal_covariance"]`.
+- `fisher_method="waveform"`: the waveform Fisher matrix for gravitational-wave likelihoods,
+  including phase/time/distance-marginalised likelihoods (via the Schur complement, with
+  `marginalized_reference` on `LaplacePosteriorEstimator`) and detector-based sky frames.
+- `map_restarts` (default 4): independent `differential_evolution` runs in the MAP search, keeping
+  the best.
+- `map_spike_guard` (default True) and `map_spike_delta`: reject MAP candidates that sit on a
+  needle spike of the log-posterior.
+- `map_vectorized` (default True): evaluate the `differential_evolution` population as one batch,
+  so the MAP search uses `npool`.
+- `smc_prior_flow`, `smc_prior_fraction`, `smc_progress` and `smc_plot_every` for SMC runs.
+- `sampling_cov`: supply a precomputed covariance.
+- `cov_scaling` accepts a per-parameter dict.
+- `jacobian_cap_scale`, `hessian_kwargs`, `use_unit_cube` and `max_iterations` options.
+- Validation of the Hessian-derived covariance along its principal axes; directions wider than
+  the Gaussian predicts are inflated.
+- Checkpoint and resume for the batched resampling methods and SMC.
+- Parallel likelihood evaluation through `npool` / `pool`.
+- `bilby_laplace.comparison`: JSD/EMD comparison of results against a reference, used by the
+  examples.
+- Examples: gaussian, rosenbrock, BBH_HLV, BNS_3G and AntiGlitch.
+
+### Changed
+
+- `resample="inprior"` uses `proposal_covariance="full"` by default; every other method uses
+  `"diagonal"`.
+- The multiprocessing pool is created before the MAP search.
+- A secondary mode with a higher log-posterior than the primary MAP logs a warning.
+- `fail_on_error` defaults to True.
+- Unit-cube Hessian defaults: `initial_step=0.001`, `step_factor=2`, `maxiter=10`; the unit-cube
+  precision is floored at the prior precision.
+- Renamed to reflect that the estimated matrix is the posterior precision, not the Fisher
+  information: `matrix.py` → `laplace.py`, `FisherMatrixPosteriorEstimator` →
+  `LaplacePosteriorEstimator`, `calculate_FIM` → `calculate_posterior_precision`, `calculate_iFIM`
+  → `calculate_posterior_covariance`.
+- The `sample_array`/`sample_dataframe` sentinel for the expansion point is `sample="MAP"` (was
+  `"maxL"`); other strings raise `ValueError`.
+
+### Fixed
+
+- `Constraint` priors are applied in the MAP search, Hessian, mode search and Laplace evidence
+  (`log_prior` delegates to `PriorDict.ln_prob`; MAP starts use `sample_subset_constrained`).
+- `log_evidence_laplace` includes bilby's constraint normalisation factor.
+- `prior_parameters` substitution happens before the prior-support test in the SMC and in-prior
+  draws.
+- Latin hypercube starts that violate a constraint are dropped from the mode search.
+- The hypercube mode search keeps the best polish of a region, not the first.
+- An exact symmetry mirror is not rejected as a duplicate of its own source by the separation gate.
+- The waveform Fisher uses the injection as the marginalised-parameter reference only when
+  `use_injection_for_map=True`.
+- Local MAP polishes wrap periodic parameters instead of treating their bounds as walls.
+- `resample=None` reports the fraction of draws outside the prior support.
+- The sampling seed reaches every stage, including SMC.
+
+### Removed
+
+- The manual finite-difference Hessian fallback and `fd_eps`; `scipy>=1.15` is required.
+- `get_maximum_likelihood_sample()`, an alias of `get_MAP_sample()`.
+- The bounds arguments of `_pool_log_likelihood`; support is tested once per batch.
+- The GW150914 example (needs a data download).
+
+---
+
+## [0.1.0] — 2026-03-02
+
+Initial release.
+
+### Added
+
+- `Fisher` sampler class — a Bilby plugin that registers as `bilby.laplace` /
+  `laplace` via the `bilby.samplers` entry-point group.
+- `FisherMatrixPosteriorEstimator` — computes the MAP estimate and inverse Fisher
+  Information Matrix (iFIM) covariance for a given likelihood and prior.
+- FIM calculation using `scipy.differentiate.hessian` (scipy ≥ 1.15) with automatic
+  finite-difference fallback for older scipy.
+- Rejection sampling and importance sampling resampling modes.
+- Batched sampling loop with a `tqdm` progress bar reporting per-batch efficiency.
+- Optional corner diagnostic plot (`plot_diagnostic=True`) comparing the Gaussian
+  proposal with the resampled posterior.
+- `result.meta_data["run_statistics"]` populated with sampling efficiency, likelihood
+  evaluation count, and wall-clock time.
+- `examples/gaussian_example.py` — minimal runnable demo on a 2-D Gaussian likelihood.
+- Version managed by `setuptools-scm` from git tags.
+
+[Unreleased]: https://github.com/GregoryAshton/bilby-laplace/compare/v0.2.0...HEAD
+[0.2.0]: https://github.com/GregoryAshton/bilby-laplace/compare/v0.1.0...v0.2.0
+[0.1.0]: https://github.com/GregoryAshton/bilby-laplace/releases/tag/v0.1.0

bilby_laplace-0.2.0/CITATION.cff

@@ -0,0 +1,21 @@
+cff-version: 1.2.0
+message: "If you use this software, please cite it as below."
+title: "bilby-laplace"
+abstract: >-
+  A Bilby sampler plugin that builds a Laplace-approximation mixture (MAP
+  search, secondary-mode search, curvature at each mode) and uses it to seed
+  or correct stochastic sampling.
+type: software
+authors:
+  - family-names: Ashton
+    given-names: Gregory
+license: MIT
+repository-code: "https://github.com/GregoryAshton/bilby-laplace"
+url: "https://gregoryashton.github.io/bilby-laplace/"
+keywords:
+  - bayesian inference
+  - laplace approximation
+  - bilby
+  - gravitational waves
+version: 0.2.0
+date-released: 2026-11-11

bilby_laplace-0.2.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Gregory Ashton
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

bilby_laplace-0.2.0/MANIFEST.in

@@ -0,0 +1,6 @@
+# setuptools-scm includes every tracked file in the sdist; keep it to the
+# package, tests and docs. The example scripts stay in the repository and the
+# Zenodo archive.
+prune examples
+prune .github
+exclude .pre-commit-config.yaml .flake8 mkdocs.yml

bilby_laplace-0.2.0/PKG-INFO

@@ -0,0 +1,131 @@
+Metadata-Version: 2.4
+Name: bilby-laplace
+Version: 0.2.0
+Summary: Bilby sampler plugin: Laplace approximation with rejection/importance resampling
+Author: Gregory Ashton
+License-Expression: MIT
+Project-URL: Homepage, https://github.com/GregoryAshton/bilby-laplace
+Project-URL: Documentation, https://gregoryashton.github.io/bilby-laplace/
+Project-URL: Source, https://github.com/GregoryAshton/bilby-laplace
+Project-URL: Issues, https://github.com/GregoryAshton/bilby-laplace/issues
+Project-URL: Changelog, https://github.com/GregoryAshton/bilby-laplace/blob/main/CHANGELOG.md
+Keywords: bayesian inference,laplace approximation,bilby,gravitational waves
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Science/Research
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Scientific/Engineering :: Astronomy
+Classifier: Topic :: Scientific/Engineering :: Physics
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: bilby>=2.7
+Requires-Dist: numpy
+Requires-Dist: scipy>=1.15
+Requires-Dist: pandas
+Requires-Dist: tqdm
+Provides-Extra: docs
+Requires-Dist: mkdocs<2; extra == "docs"
+Requires-Dist: mkdocs-material; extra == "docs"
+Requires-Dist: mkdocstrings[python]; extra == "docs"
+Requires-Dist: pymdown-extensions; extra == "docs"
+Provides-Extra: test
+Requires-Dist: pytest; extra == "test"
+Requires-Dist: pytest-cov; extra == "test"
+Dynamic: license-file
+
+# bilby-laplace
+
+A [Bilby](https://bilby-dev.github.io/bilby/) sampler plugin that estimates posteriors
+via the **Laplace approximation** — a Gaussian fitted at the maximum a posteriori (MAP)
+point using the Hessian of the log-posterior — followed by optional resampling to
+correct for non-Gaussianity.
+
+The method is fast, scales well to moderate dimensions, and produces asymptotically
+exact posterior samples when the true posterior is close to Gaussian. It is
+useful as a cheap cross-check against nested sampling results.
+
+> **Documentation:** full guides, the configuration/API reference, and background on
+> the method are at **<https://gregoryashton.github.io/bilby-laplace/>**.
+
+Originally derived from [bilby PR #933](https://github.com/bilby-dev/bilby/pull/933).
+
+## Installation
+
+```bash
+pip install bilby-laplace
+```
+
+Or from source:
+
+```bash
+git clone https://github.com/GregoryAshton/bilby-laplace
+cd bilby-laplace
+pip install -e .
+```
+
+Bilby discovers the sampler automatically via its plugin entry-point system — no
+further configuration is needed.
+
+## Quick start
+
+```python
+import bilby
+
+result = bilby.run_sampler(
+    likelihood=likelihood,
+    priors=priors,
+    sampler="laplace",
+    outdir="outdir",
+    label="my_run",
+)
+
+result.plot_corner()
+print(result.posterior)
+```
+
+## Features
+
+- Several resampling strategies to correct for non-Gaussianity: `rejection`,
+  `importance`, `inprior`, `smc` (via [aspire](https://github.com/bilby-dev/aspire)),
+  or none.
+- Two covariance routes: the numerical Hessian (`fisher_method="hessian"`, any
+  likelihood) or the genuine waveform Fisher matrix (`fisher_method="waveform"`, for
+  gravitational-wave likelihoods).
+- Supply a precomputed covariance via `sampling_cov` (e.g. from gwfast / GWFish).
+- Multimodal posteriors via SMC with `n_modes > 1`.
+- Laplace log-evidence always available; rejection and SMC add independent estimates.
+
+See the [documentation](https://gregoryashton.github.io/bilby-laplace/) for the full list
+of options and guidance on choosing between them.
+
+## Examples
+
+Runnable examples are in [`examples/`](https://github.com/GregoryAshton/bilby-laplace/tree/main/examples) (gaussian, rosenbrock, BBH, BNS).
+Each has a `Makefile`:
+
+```bash
+cd examples/gaussian
+make laplace
+make compare
+```
+
+## Citing
+
+If you use `bilby-laplace`, please cite the software release; citation metadata
+is in [`CITATION.cff`](https://github.com/GregoryAshton/bilby-laplace/blob/main/CITATION.cff) (GitHub's "Cite this repository" button
+reads it).
+
+## License
+
+MIT; see [`LICENSE`](https://github.com/GregoryAshton/bilby-laplace/blob/main/LICENSE).
+
+## Documentation development
+
+```bash
+pip install -e ".[docs]"
+mkdocs serve
+```

bilby_laplace-0.2.0/README.md

@@ -0,0 +1,92 @@
+# bilby-laplace
+
+A [Bilby](https://bilby-dev.github.io/bilby/) sampler plugin that estimates posteriors
+via the **Laplace approximation** — a Gaussian fitted at the maximum a posteriori (MAP)
+point using the Hessian of the log-posterior — followed by optional resampling to
+correct for non-Gaussianity.
+
+The method is fast, scales well to moderate dimensions, and produces asymptotically
+exact posterior samples when the true posterior is close to Gaussian. It is
+useful as a cheap cross-check against nested sampling results.
+
+> **Documentation:** full guides, the configuration/API reference, and background on
+> the method are at **<https://gregoryashton.github.io/bilby-laplace/>**.
+
+Originally derived from [bilby PR #933](https://github.com/bilby-dev/bilby/pull/933).
+
+## Installation
+
+```bash
+pip install bilby-laplace
+```
+
+Or from source:
+
+```bash
+git clone https://github.com/GregoryAshton/bilby-laplace
+cd bilby-laplace
+pip install -e .
+```
+
+Bilby discovers the sampler automatically via its plugin entry-point system — no
+further configuration is needed.
+
+## Quick start
+
+```python
+import bilby
+
+result = bilby.run_sampler(
+    likelihood=likelihood,
+    priors=priors,
+    sampler="laplace",
+    outdir="outdir",
+    label="my_run",
+)
+
+result.plot_corner()
+print(result.posterior)
+```
+
+## Features
+
+- Several resampling strategies to correct for non-Gaussianity: `rejection`,
+  `importance`, `inprior`, `smc` (via [aspire](https://github.com/bilby-dev/aspire)),
+  or none.
+- Two covariance routes: the numerical Hessian (`fisher_method="hessian"`, any
+  likelihood) or the genuine waveform Fisher matrix (`fisher_method="waveform"`, for
+  gravitational-wave likelihoods).
+- Supply a precomputed covariance via `sampling_cov` (e.g. from gwfast / GWFish).
+- Multimodal posteriors via SMC with `n_modes > 1`.
+- Laplace log-evidence always available; rejection and SMC add independent estimates.
+
+See the [documentation](https://gregoryashton.github.io/bilby-laplace/) for the full list
+of options and guidance on choosing between them.
+
+## Examples
+
+Runnable examples are in [`examples/`](https://github.com/GregoryAshton/bilby-laplace/tree/main/examples) (gaussian, rosenbrock, BBH, BNS).
+Each has a `Makefile`:
+
+```bash
+cd examples/gaussian
+make laplace
+make compare
+```
+
+## Citing
+
+If you use `bilby-laplace`, please cite the software release; citation metadata
+is in [`CITATION.cff`](https://github.com/GregoryAshton/bilby-laplace/blob/main/CITATION.cff) (GitHub's "Cite this repository" button
+reads it).
+
+## License
+
+MIT; see [`LICENSE`](https://github.com/GregoryAshton/bilby-laplace/blob/main/LICENSE).
+
+## Documentation development
+
+```bash
+pip install -e ".[docs]"
+mkdocs serve
+```

bilby_laplace-0.2.0/docs/background/theory.md

@@ -0,0 +1,93 @@
+# Fisher, Hessian & precision
+
+A common point of confusion: the matrix the default route computes is **not** the
+Fisher information matrix, even though such code is often loosely called "the Fisher
+matrix". This page makes the distinction precise, because it explains both the
+naming in the code and why the `"waveform"` route can give a better-behaved result.
+
+## The decomposition
+
+For a Gaussian likelihood with mean \( \mu(\theta) \), data \( d \), and noise
+covariance \( C \), write the residual \( r = d - \mu(\theta) \). The second
+derivative of the log-likelihood splits cleanly:
+
+\[
+-\partial_i \partial_j \log L
+  = \underbrace{(\partial_i \mu)^{\mathsf T} C^{-1} (\partial_j \mu)}_{\textstyle F_{ij}\ \text{(Fisher)}}
+  \;-\; \underbrace{r^{\mathsf T} C^{-1} (\partial_i \partial_j \mu)}_{\textstyle \text{residual}\times\text{curvature}}
+\]
+
+For gravitational waves the noise-weighted inner product
+\( (a\mid b) = 4\,\mathrm{Re}\int a^* b / S_n\, \mathrm{d}f \) **is** \( a^{\mathsf T} C^{-1} b \),
+so with \( \mu \to h(\theta) \):
+
+\[
+-\partial_i \partial_j \log L = (\partial_i h \mid \partial_j h) - (r \mid \partial_i \partial_j h).
+\]
+
+Three distinct objects appear:
+
+| Object | Definition | Properties |
+|---|---|---|
+| **Fisher information** \( F \) | \( \mathbb{E}[-\partial^2 \log L] = (\partial_i h\mid\partial_j h) \) | expectation; model + noise only; positive semi-definite; data-independent |
+| **Observed information** \( J \) | \( -\partial^2 \log L \) at a point | single realisation; \( = F - (r\mid\partial^2 h) \); can be indefinite |
+| **Posterior precision** | \( -\partial^2 \log(L\pi) = J + (-\partial^2\log\pi) \) | what the Laplace approximation actually uses |
+
+## Why the routes differ
+
+- **`fisher_method="hessian"`** numerically estimates the *posterior precision* —
+  the third row, including the prior. It is the observed information plus the prior
+  curvature.
+- **`fisher_method="waveform"`** computes \( F \) directly — the first term only —
+  plus a diagonal prior precision.
+
+The two agree **in expectation**: at the true parameters the residual is pure noise,
+\( \mathbb{E}[(r\mid X)] = 0 \), so \( \mathbb{E}[-\partial^2\log L] = F \). This is
+the Fisher information identity, and the basis of the Cramér–Rao bound.
+
+For a **single** dataset they differ by the residual–curvature term, which is
+zero-mean but has nonzero variance, is **not sign-definite**, and shrinks roughly
+like \( \mathcal{O}(1/\mathrm{SNR}) \). Dropping it is the *linear-signal
+approximation*. Evaluating at the MAP does not remove it: the gradient vanishes
+there, but the residual still has a component off the signal manifold that projects
+onto \( \partial_i\partial_j h \).
+
+## Consequences in practice
+
+1. The Hessian route computes the **noisy** object: even with perfect
+   differentiation, \( J \) includes the indefinite residual–curvature term, which is
+   why the inversion needs eigenvalue flooring. The waveform route computes the clean
+   PSD \( F \).
+2. The Hessian route differentiates a **scalar** to get a *second* derivative — the
+   accuracy ceiling is roughly \( \sqrt{\varepsilon_f} \). The Fisher needs only
+   *first* derivatives of \( h \), then a Gram product — better conditioned and
+   positive semi-definite for free.
+
+A subtlety worth stating: for a *specific* dataset, the inverse observed information
+\( J^{-1} \) is actually the better local Gaussian fit to the realised posterior (the
+Laplace approximation *is* a Taylor expansion of the realised log-posterior). Tools
+like [gwfast](https://github.com/CosmoStatGW/gwfast) and
+[GWFish](https://github.com/janosch314/GWFish) compute \( F \) because they are doing
+*forecasting* (expected errors, Cramér–Rao). Here we only need a **proposal**
+covariance — and \( F \) is a clean, PSD, well-conditioned surrogate that equals
+\( J \) in expectation, so it is usually the better engineering choice when available.
+
+## Conditioning
+
+GW Fisher matrices are notoriously ill-conditioned — condition numbers of
+\( 10^{10} \)–\( 10^{20} \) from strong degeneracies (distance–inclination, the
+masses, sky position). Inverting in double precision loses about
+\( \log_{10}(\kappa) \) digits, so a poorly-conditioned matrix can be inverted to
+*zero* correct digits.
+
+Two independent sources contribute:
+
+- **Scale-driven** conditioning (parameters span many orders of magnitude). Removed
+  by **diagonal preconditioning** — rescaling by \( \sqrt{\mathrm{diag}} \) before
+  inversion. This is the shared trick used by both GWFish and gwfast.
+- **Correlation-driven** conditioning (genuine parameter degeneracies). Preconditioning
+  cannot help here; only a better parameterisation or the Fisher form does.
+
+The unit-cube Hessian path already normalises scales via the prior CDF transform, so
+its residual conditioning is correlation-driven; the parameter-space and
+waveform-Fisher paths benefit much more from preconditioning.

bilby_laplace-0.2.0/docs/changelog.md

@@ -0,0 +1 @@
+--8<-- "CHANGELOG.md"

bilby_laplace-0.2.0/docs/examples.md

@@ -0,0 +1,51 @@
+# Examples
+
+Runnable example scripts live in the [`examples/`](https://github.com/GregoryAshton/bilby-laplace/tree/main/examples)
+directory. Each script supports multiple samplers and a `--compare` mode that prints
+a summary table and a corner plot overlaying the methods.
+
+Each example directory has a `Makefile`; run targets from within it:
+
+```bash
+cd examples/gaussian
+make laplace          # raw Laplace (no resampling)
+make rejection        # rejection resampling
+make rejection-user   # rejection with a user-supplied sampling_cov
+make smc              # SMC resampling
+make dynesty          # nested-sampling reference
+make compare          # load all results, print evidence table, plot
+```
+
+## Available examples
+
+| Example | What it demonstrates |
+|---|---|
+| `examples/gaussian` | 2-D correlated Gaussian — the Laplace approximation is exact here, so it is the cleanest correctness check. Includes a `rejection-user` variant that feeds the known covariance via `sampling_cov`. |
+| `examples/rosenbrock` | A strongly non-Gaussian, curved target — a stress test where `resample="smc"` matters. |
+| `examples/BBH_HLV` | A simulated binary-black-hole injection in an H–L–V network (no data download needed). A realistic high-dimensional GW problem. |
+| `examples/BNS_3G` | A binary-neutron-star injection in a 3rd-generation network. |
+| `examples/AntiGlitch` | A simulated anti-glitch in Gaussian noise. |
+
+## A minimal script
+
+```python
+import bilby
+
+likelihood = ...   # any bilby Likelihood
+priors = ...       # a bilby PriorDict
+
+result = bilby.run_sampler(
+    likelihood=likelihood,
+    priors=priors,
+    sampler="laplace",
+    resample="rejection",
+    target_nsamples=5000,
+    plot_diagnostic=True,
+    outdir="outdir",
+    label="example",
+)
+result.plot_corner()
+```
+
+See the [gaussian example](https://github.com/GregoryAshton/bilby-laplace/blob/main/examples/gaussian/run.py)
+for a complete, self-contained script including the likelihood definition.