Posterior Matching for Arbitrary Conditioning
Posterior Matching for Arbitrary Conditioning
复制标题
DOI：
--
复制
发表时间：
2022-01
期刊：
影响因子：
--
通讯作者：
R. Strauss;Junier B. Oliva
中科院分区：
文献类型：
--
作者：
R. Strauss;Junier B. Oliva
任意条件是无监督学习中的一个重要问题，我们寻求对某些数据下的条件密度 $p(\mathbf{x}_u \mid \mathbf{x}_o)$ 进行建模，对于所有可能的非相交子集 $o, u \subset \{1, \dots , d\}$。然而，绝大多数密度估计仅关注于对联合分布 $p(\mathbf{x})$ 进行建模，其中特征之间的重要条件依赖关系是不透明的。我们提出了一个简单而通用的框架，即后验匹配，它使变分自动编码器（VAE）能够执行任意调节，而无需修改 VAE 本身。后验匹配适用于许多现有的基于 VAE 的联合密度估计方法，从而规避了以前的任意调节方法所需的专用模型。我们发现，对于具有各种 VAE（例如离散、分层、VaDE）的各种任务，后验匹配可以与当前最先进的方法相媲美或优于当前最先进的方法。
Arbitrary conditioning is an important problem in unsupervised learning, where we seek to model the conditional densities $p(\mathbf{x}_u \mid \mathbf{x}_o)$ that underly some data, for all possible non-intersecting subsets $o, u \subset \{1, \dots , d\}$. However, the vast majority of density estimation only focuses on modeling the joint distribution $p(\mathbf{x})$, in which important conditional dependencies between features are opaque. We propose a simple and general framework, coined Posterior Matching, that enables Variational Autoencoders (VAEs) to perform arbitrary conditioning, without modification to the VAE itself. Posterior Matching applies to the numerous existing VAE-based approaches to joint density estimation, thereby circumventing the specialized models required by previous approaches to arbitrary conditioning. We find that Posterior Matching is comparable or superior to current state-of-the-art methods for a variety of tasks with an assortment of VAEs (e.g.~discrete, hierarchical, VaDE).