arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1206.4602math.NAcs.LGcs.NAstat.ML
收藏

拟牛顿方法:一个新方向

Quasi-Newton Methods: A New Direction

Max Planck Institute for Intelligent Systems(马克斯·普朗克智能系统研究所)

Philipp Hennig, Martin Kiefel

AI总结:

本文通过将拟牛顿方法解释为贝叶斯线性回归的近似,揭示了经典算法的缺陷,并提出了一种新的非参数拟牛顿方法,在相似计算成本下更高效地利用信息。

摘要:

在拟牛顿方法发明四十年后,它们仍然是无约束数值优化中的最先进技术。虽然通常不被这样解释,但这些是拟合目标函数的局部二次逼近的学习算法。我们表明,许多(包括最流行的)拟牛顿方法可以解释为在不同先验假设下贝叶斯线性回归的近似。这一新概念阐明了经典算法的一些缺陷,并为一种新颖的非参数拟牛顿方法指明了道路,该方法能够在与之前方法相似的计算成本下更有效地利用可用信息。

英文摘要:

Four decades after their invention, quasi-Newton methods are still state of the art in unconstrained numerical optimization. Although not usually interpreted thus, these are learning algorithms that fit a local quadratic approximation to the objective function. We show that many, including the most popular, quasi-Newton methods can be interpreted as approximations of Bayesian linear regression under varying prior assumptions. This new notion elucidates some shortcomings of classical algorithms, and lights the way to a novel nonparametric quasi-Newton method, which is able to make more efficient use of available information at computational cost similar to its predecessors.

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。

↑