Nonlinear Stein Variational Gradient Descent for Learning Diversified Mixture Models-猫眼课题宝

DOI：

发表时间：

2019-05

期刊：

影响因子：

通讯作者：

中科院分区：

其他

文献类型：

作者：

Dilin Wang;Qiang Liu

多样化已被证明是一个强大的机制，学习鲁棒模型在非凸设置。一个值得注意的例子是学习混合模型，在这种模型中，不同混合成分之间的多样性使我们能够防止模型崩溃现象，并从观察到的数据中捕获更多的模式。在这项工作中，我们提出了一个变分的方法多样性促进学习，它利用熵功能作为一个自然的机制，执行多样性。我们开发了一个简单而有效的功能梯度为基础的算法优化变分目标函数，它提供了一个显着的推广斯坦变分梯度下降（SVGD）。我们在各种具有挑战性的真实的世界问题上测试我们的方法，包括深度嵌入式聚类和深度异常检测。实验结果表明，该方法为多样性促进学习提供了一种有效的机制，与现有方法相比，取得了实质性的改进.

Diversification has been shown to be a powerful mechanism for learning robust models in nonconvex settings. A notable example is learning mixture models, in which enforcing diversity between the different mixture components allows us to prevent the model collapsing phenomenon and capture more patterns from the observed data. In this work, we present a variational approach for diversity-promoting learning, which leverages the entropy functional as a natural mechanism for enforcing diversity. We develop a simple and efficient functional gradientbased algorithm for optimizing the variational objective function, which provides a significant generalization of Stein variational gradient descent (SVGD). We test our method on various challenging real world problems, including deep embedded clustering and deep anomaly detection. Empirical results show that our method provides an effective mechanism for diversitypromoting learning, achieving substantial improvement over existing methods.

Nonlinear Stein Variational Gradient Descent for Learning Diversified Mixture Models

文献摘要

相似文献