Importance nested sampling with normalising flows

Abstract We present an improved version of the nested sampling algorithm nessai in which core is modified to use importance weights. In algorithm, samples are drawn from a mixture normalising flows and requirement for be independently identically distributed (i.i.d.) according prior relaxed. Furthermore, it allows added any order, likelihood constraint, evidence updated with batches samples. call i-nessai . first validate using analytic likelihoods known Bayesian evidences show that estimates unbiased up 32 dimensions. compare standard Rosenbrock likelihood, results consistent whilst producing more precise estimates. then test on 64 simulated gravitational-wave signals binary black hole coalescence produces parameters. our those obtained dynesty find requires 2.68 13.3 times fewer evaluations converge, respectively. also 80 s neutron star signal reduced-order-quadrature basis that, average, converges 24 min, only requiring 1.01 × 10 6 compared $1.42 1.42 $4.30 10^{7}$?> 4.30 7 These demonstrate being efficient.