starling.structure.bme_utils.BMEResult

Container for BME optimization results.

Methods

`__init__`
`diagnostics`	Diagnose BME reweighting results and identify potential issues.
`print_diagnostics`	Print a formatted diagnostic report for this BME result.

Attributes

`kl_divergence`	returns the KL divergence (relative entropy) from phi.
`weights`
`initial_weights`
`lambdas`
`chi_squared_initial`
`chi_squared_final`
`phi`
`n_iterations`
`success`
`message`
`theta`
`observables`
`calculated_values`
`metadata`

property kl_divergence: float: returns the KL divergence (relative entropy) from phi.

diagnostics(warn_threshold: float = 0.5) → dict[source]

Diagnose BME reweighting results and identify potential issues.

Notes

We report two notions of effective sample size:

neff_entropy (N_eff^(S)): entropy-based, derived from Φ = exp(-D_KL). This is the standard BME measure: N_eff^(S) = N * Φ.
neff_renyi2 (N_eff^(2)): 1 / sum_i w_i^2 (Rényi-2 / participation ratio). This is more sensitive to a few large weights, so it is always <= neff_entropy for the same weights.

print_diagnostics(warn_threshold: float = 0.5)[source]: Print a formatted diagnostic report for this BME result.

__init__(weights: ndarray, initial_weights: ndarray, lambdas: ndarray, chi_squared_initial: float, chi_squared_final: float, phi: float, n_iterations: int, success: bool, message: str, theta: float, observables: List[ExperimentalObservable], calculated_values: ndarray, metadata: dict) → None