Interface DlmMleResult

Result from MLE estimation with JS-idiomatic names.

interface DlmMleResult {
    arCoefficients?: number[];
    compilationMs: number;
    deviance: number;
    devianceHistory: number[];
    elapsed: number;
    fit: DlmFitResult;
    iterations: number;
    obsStd: number;
    priorPenalty?: number;
    processStd: number[];
}

Properties

arCoefficients?: number[]

Estimated AR coefficients (only when fitAr=true). In MATLAB DLM, this is arphi.

compilationMs: number

Wall-clock time in ms for the first optimizer step (JIT compilation + one gradient pass). In MATLAB DLM, this is jitMs.

deviance: number

Deviance: -2 · log-likelihood at optimum. In MATLAB DLM, this is lik.

devianceHistory: number[]

Optimization history: deviance at each iteration. In MATLAB DLM, this is likHistory.

elapsed: number

Wall-clock time in ms (total: setup + all iterations + final dlmFit)

Full DLM fit result using the estimated parameters

iterations: number

Number of optimizer iterations

obsStd: number

Estimated observation noise std dev. In MATLAB DLM, this is s.

priorPenalty?: number

Prior penalty at the optimum: −2·log p(θ*). Only present when a custom loss function is used.

processStd: number[]

Estimated state noise std devs (diagonal of √W). In MATLAB DLM, this is w.