HINT: Hierarchical Invertible Neural Transport for Density Estimation and Bayesian Inference
HINT: Hierarchical Invertible Neural Transport for Density Estimation and Bayesian Inference
复制标题
提示：用于密度估计和贝叶斯推理的分层可逆神经传输
DOI：
10.1609/aaai.v35i9.16997
复制
发表时间：
2019
期刊：
影响因子：
--
通讯作者：
U. Köthe
中科院分区：
文献类型：
--
作者：
Jakob Kruse;Gianluca Detommaso;Robert Scheichl;U. Köthe
最近的许多可逆神经结构都是基于耦合块设计，其中变量被分为两个子集，作为易于可逆（通常是仿射）三角变换的输入。虽然这样的变换是可逆的，但它的雅可比矩阵非常稀疏，因此可能缺乏表达性。这项工作提出了一个简单的补救办法，即注意到细分和（仿射）耦合
Many recent invertible neural architectures are based on coupling block designs where variables are divided in two subsets which serve as inputs of an easily invertible (usually affine) triangular transformation. While such a transformation is invertible, its Jacobian is very sparse and thus may lack expressiveness. This work presents a simple remedy by noting that subdivision and (affine) coupling
can be repeated recursively within the resulting subsets, leading to an efficiently invertible block with dense, triangular Jacobian. By formulating our recursive coupling scheme via a hierarchical architecture, HINT allows sampling from a joint distribution p(y,x) and the corresponding posterior p(x|y) using a single invertible network. We evaluate our method on some standard data sets and benchmark its full power for density estimation and Bayesian inference on a novel data set of 2D shapes in Fourier parameterization, which enables consistent visualization of samples for different dimensionalities.
DOI：
--
发表时间：
2019-07
期刊：
ArXiv
影响因子：
--
作者：
Yang Song;Chenlin Meng;Stefano Ermon
通讯作者：
Yang Song;Chenlin Meng;Stefano Ermon