arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1507.04513stat.ML
收藏

基于期望传播的可扩展高斯过程分类

Scalable Gaussian Process Classification via Expectation Propagation

Daniel Hernández-Lobato, José Miguel Hernández-Lobato

摘要:

变分方法最近被认为是用于将高斯过程分类器的训练过程扩展到大数据集。作为替代方案,我们在这里描述如何使用期望传播有效地训练这些分类器。所提出的方法允许处理具有数百万个数据实例的数据集。更确切地说,它可以用于(i)以分布式方式进行训练,其中数据实例被发送到执行所需计算的不同节点,以及用于(ii)使用梯度的随机逼近来最大化边际似然的估计。几个实验表明,所描述的方法是有竞争力的变分方法。

英文摘要:

Variational methods have been recently considered for scaling the training process of Gaussian process classifiers to large datasets. As an alternative, we describe here how to train these classifiers efficiently using expectation propagation. The proposed method allows for handling datasets with millions of data instances. More precisely, it can be used for (i) training in a distributed fashion where the data instances are sent to different nodes in which the required computations are carried out, and for (ii) maximizing an estimate of the marginal likelihood using a stochastic approximation of the gradient. Several experiments indicate that the method described is competitive with the variational approach.

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。

↑