We establish $L_q$ convergence for Hamiltonian Monte Carlo algorithms. More specifically, under mild conditions for the associated Hamiltonian motion, we show that the outputs of the algorithms converge (strongly for $2\le q<\infty$ and weakly for $1<q<2$) to the desired target distribution.
翻译:我们为汉密尔顿·蒙特卡洛算法建立了$L_q美元趋同法。更具体地说,在相关的汉密尔顿动议的温和条件下,我们证明算法的产出(主要是2美元,主要是2美元,小于1美元,小于1美元)与预期的目标分配一致。