jesterTOV.inference.likelihoods.mock_mr.MockMassRadiusLikelihood#

class MockMassRadiusLikelihood(
psr_name,
mean_mass,
mean_radius,
std_mass,
std_radius,
correlation,
penalty_value=0.0,
N_masses_evaluation=100,
N_masses_batch_size=20,
seed=42,
)[source]#

Bases: LikelihoodBase

Likelihood for a synthetic bivariate Gaussian mass-radius measurement.

The “observation” is a 2-D Gaussian centred on (mean_mass, mean_radius) with standard deviations std_mass and std_radius and Pearson correlation correlation. The log-likelihood for a model M-R curve is estimated by Monte-Carlo integration: masses are pre-sampled from the joint distribution at initialisation via MultivariateGaussianPrior, the EOS is used to interpolate the predicted radius at each sampled mass, and the bivariate normal log-pdf is evaluated and averaged.

The covariance matrix is built from the given standard deviations and correlation as

\[\begin{split}\Sigma = \begin{pmatrix} \sigma_M^2 & \rho\,\sigma_M\sigma_R \\ \rho\,\sigma_M\sigma_R & \sigma_R^2 \end{pmatrix}\end{split}\]
Parameters:
  • psr_name (str) – Identifier for this measurement (e.g., "PSR0"). Used only for logging and bookkeeping.

  • mean_mass (float) – Mean of the observed mass distribution (\(M_\odot\)).

  • mean_radius (float) – Mean of the observed radius distribution (km).

  • std_mass (float) – Standard deviation of the mass measurement (\(M_\odot\)).

  • std_radius (float) – Standard deviation of the radius measurement (km).

  • correlation (float) – Pearson correlation coefficient between mass and radius, \(\rho \in (-1, 1)\).

  • penalty_value (float, optional) – Log-likelihood penalty applied when the pre-sampled mass exceeds \(M_\mathrm{TOV}\) (default: 0.0, i.e. no penalty).

  • N_masses_evaluation (int, optional) – Number of (mass, radius) samples drawn at initialisation for Monte-Carlo integration (default: 100).

  • N_masses_batch_size (int, optional) – Batch size passed to jax.lax.map() for memory-efficient processing of the mass samples (default: 20).

  • seed (int, optional) – Random seed used for the deterministic mass pre-sampling (default: 42).

__init__(
psr_name,
mean_mass,
mean_radius,
std_mass,
std_radius,
correlation,
penalty_value=0.0,
N_masses_evaluation=100,
N_masses_batch_size=20,
seed=42,
)[source]#

Methods

__init__(psr_name, mean_mass, mean_radius, ...)

evaluate(params)

Evaluate the log likelihood for given EOS parameters.

Attributes

data

The data for the likelihood.

model

The model for the likelihood.

psr_name

penalty_value

N_masses_evaluation

N_masses_batch_size

seed

distribution

fixed_mass_samples

N_masses_batch_size: int#
N_masses_evaluation: int#
distribution: MultivariateGaussianPrior#
evaluate(params)[source]#

Evaluate the log likelihood for given EOS parameters.

For each pre-sampled mass the corresponding radius is obtained by linear interpolation of the EOS M-R curve. The bivariate normal log-pdf is computed via log_prob() and the Monte-Carlo average is taken via the log-sum-exp trick.

Parameters:

params (dict) –

Must contain:

  • 'masses_EOS': 1-D array of neutron star masses (\(M_\odot\)) from the EOS M-R relation, sorted in ascending order.

  • 'radii_EOS': 1-D array of corresponding radii (km).

Return type:

Float

Returns:

Float – Log likelihood \(\ln \mathcal{L}\).

fixed_mass_samples: Float[Array, 'n_samples']#
penalty_value: float#
psr_name: str#
seed: int#