arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2104.02865cs.LGcs.NAmath.NAstat.ML
收藏

变分贝叶斯的拟牛顿拟蒙特卡罗方法

Quasi-Newton Quasi-Monte Carlo for variational Bayes

Stanford University(斯坦福大学)

Sifan Liu, Art B. Owen

摘要:

许多机器学习问题优化的目标必须用噪声来衡量。主要的方法是一阶随机梯度下降,在每一步使用一个或多个蒙特卡罗(MC)样本。在某些情况下,病态使二阶方法(如L-BFGS)更有效。我们研究使用随机准蒙特卡罗(RQMC)抽样这样的问题。当MC采样的均方根误差(RMSE)为$O(n^{-1/2})$时,则RQMC的RMSE为$o(n^{-1/2})$,在有利的设置中,该RMSE可以接近$O(n^{-3/2})$。我们证明,改进的采样精度直接转化为改进的优化。在我们对变分贝叶斯的实证研究中,使用随机L-BFGS的RQMC大大加快了优化速度,有时会找到比MC更好的参数值。

英文摘要:

Many machine learning problems optimize an objective that must be measured with noise. The primary method is a first order stochastic gradient descent using one or more Monte Carlo (MC) samples at each step. There are settings where ill-conditioning makes second order methods such as L-BFGS more effective. We study the use of randomized quasi-Monte Carlo (RQMC) sampling for such problems. When MC sampling has a root mean squared error (RMSE) of $O(n^{-1/2})$ then RQMC has an RMSE of $o(n^{-1/2})$ that can be close to $O(n^{-3/2})$ in favorable settings. We prove that improved sampling accuracy translates directly to improved optimization. In our empirical investigations for variational Bayes, using RQMC with stochastic L-BFGS greatly speeds up the optimization, and sometimes finds a better parameter value than MC does.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑