The Apogee to Apogee Path Sampler

Among Markov chain Monte Carlo algorithms, Hamiltonian (HMC) is often the algorithm of choice for complex, high-dimensional target distributions; however, its efficiency notoriously sensitive to integration-time tuning parameter. When integrating both forward and backward in time using same leapfrog integration step as HMC, set apogees, local maxima potential along a path, whatever point (position momentum) path chosen initialize integration. We present Apogee Path Sampler (AAPS), which uses this invariance create simple yet generic methodology constructing proposing from it accepting or rejecting that proposal so intended distribution. demonstrate empirically AAPS has similar HMC but much more robust setting equivalent parameter, number apogees crosses. Supplementary materials article are available online.