Function gibbsSampling

gibbsSampling(
    cpts: Map<string, CPT>,
    nodeNames: string[],
    parents: Map<string, string[]>,
    query: string,
    evidence: Evidence,
    options?: {
        burnIn?: number;
        iterations?: number;
        seed?: number;
        thin?: number;
    },
): { acceptanceRate: number; posterior: Map<number, number> }
Gibbs Sampling — Markov Chain Monte Carlo for approximate inference.

Sequentially samples each non-evidence variable from its full conditional distribution given its Markov blanket. Uses burn-in + thinning.
Parameters
- cpts: Map<string, CPT>
  — CPTs
- nodeNames: string[]
  — topological order
- parents: Map<string, string[]>
  — parent mapping
- query: string
  — query variable
- evidence: Evidence
  — observed values
- options: { burnIn?: number; iterations?: number; seed?: number; thin?: number } = {}
  — iterations (default 10000), burnIn (default 1000), thin (default 1), seed
Returns { acceptanceRate: number; posterior: Map<number, number> }
posterior distribution with acceptance rate
- Defined in pipeline/src/infer/bayesian-network.ts:865