jaxpint.likelihood

Single-pulsar log-likelihood for JaxPINT.

Composes residuals, noise covariance, and the Woodbury solver into a differentiable, JIT-compatible log-likelihood evaluation.

The Gaussian log-likelihood is evaluated via the Woodbury matrix identity to avoid forming the full n_toas x n_toas covariance matrix; see van Haasteren et al. (2009) [1] Appendix A and Lentati et al. (2013) [2] Section II.B.

References

[1]

van Haasteren et al. (2009), “On measuring the gravitational-wave background using pulsar timing arrays”, MNRAS 395, 1005.

[2]

Lentati et al. (2013), “Hyper-efficient model-independent Bayesian method for the analysis of pulsar timing data”, PRD 87, 104021.

jaxpint.likelihood.single_pulsar_logL(toa_data, timing_model, noise_model, params, external_delay=None, external_cov=None)

Per-pulsar log-likelihood with optional external injections.

Parameters:

• toa_data (TOAData) – Pulse time-of-arrival data.

• timing_model (TimingModel) – JaxPINT timing model (delay + phase components).

• noise_model (NoiseModel) – JaxPINT noise model (white + correlated noise).

• params (ParameterVector) – Timing and noise parameters for this pulsar.

• external_delay (optional array (n_toas,)) – Pre-computed external delay in seconds (e.g., sum of CW signals). Subtracted from residuals (positive delay = later arrival).

• external_cov (optional (U, Phi) tuple) – U: (n_toas, n_basis), Phi: (n_basis,). Augments the noise covariance: C += U @ diag(Phi) @ U^T.

Returns:

logL – Log-likelihood value.

Return type:

float