Trait nuts_rs::DivergenceInfo

pub trait DivergenceInfo: AsSampleStatVec + Debug + Send {
    fn start_location(&self) -> Option<&[f64]>;
    fn end_location(&self) -> Option<&[f64]>;
    fn energy_error(&self) -> Option<f64>;
    fn end_idx_in_trajectory(&self) -> Option<i64>;
    fn start_idx_in_trajectory(&self) -> Option<i64>;
    fn logp_function_error(&self) -> Option<&dyn Error>;
}

Details about a divergence that might have occured during sampling

There are two reasons why we might observe a divergence:

• The integration error of the Hamiltonian is larger than a cutoff value or nan.
• The logp function caused a recoverable error (eg if an ODE solver failed)

Required Methods

fn start_location(&self) -> Option<&[f64]>

The position in parameter space where the diverging leapfrog started

fn end_location(&self) -> Option<&[f64]>

The position in parameter space where the diverging leapfrog ended

fn energy_error(&self) -> Option<f64>

The difference between the energy at the initial location of the trajectory and the energy at the end of the diverging leapfrog step.

This is not available if the divergence was caused by a logp function error

fn end_idx_in_trajectory(&self) -> Option<i64>

The index of the end location of the diverging leapfrog.

fn start_idx_in_trajectory(&self) -> Option<i64>

The index of the start location of the diverging leapfrog.

fn logp_function_error(&self) -> Option<&dyn Error>

Return the logp function error that caused the divergence if there was any

This is not available if the divergence was cause because of a large energy difference.

Implementors