Granziol Diego, Ru Binxin, Zohren Stefan, Dong Xiaowen, Osborne Michael, Roberts Stephen
Machine Learning Research Group, University of Oxford, Walton Well Rd, Oxford OX2 6ED, UK.
Oxford-Man Institute of Quantitative Finance, Walton Well Rd, Oxford OX2 6ED, UK.
Entropy (Basel). 2019 May 31;21(6):551. doi: 10.3390/e21060551.
Efficient approximation lies at the heart of large-scale machine learning problems. In this paper, we propose a novel, robust maximum entropy algorithm, which is capable of dealing with hundreds of moments and allows for computationally efficient approximations. We showcase the usefulness of the proposed method, its equivalence to constrained Bayesian variational inference and demonstrate its superiority over existing approaches in two applications, namely, fast log determinant estimation and information-theoretic Bayesian optimisation.
高效逼近是大规模机器学习问题的核心。在本文中，我们提出了一种新颖、稳健的最大熵算法，该算法能够处理数百个矩，并允许进行计算高效的逼近。我们展示了所提出方法的实用性，它与约束贝叶斯变分推理的等价性，并在快速对数行列式估计和信息论贝叶斯优化这两个应用中证明了其相对于现有方法的优越性。
