youtor.org
Autotuning Hamiltonian Monte Carlo
>> YOUR LINK HERE: ___ http://youtube.com/watch?v=k-RycBY_XEY
Hamiltonian Monte Carlo (HMC) is an efficient sampling algorithm that can be used, for example, to solve high-dimensional inverse problems. The actual efficiency of HMC, however, depends on a very large number of tuning parameters. This video explains how HMC can find a good set of tuning parameters all by itself while sampling. The method is based on the combination of standard HMC with a quasi-Newton method that estimates the local curvature of the parameter space on the fly. • https://doi.org/10.1093/gji/ggab270
#############################
