arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2301.04791stat.MLcs.CVcs.GRcs.LG
收藏

自注意力摊销分布投影优化用于切片 Wasserstein 点云重建

Self-Attention Amortized Distributional Projection Optimization for Sliced Wasserstein Point-Cloud Reconstruction

University of Texas at Austin(德克萨斯大学奥斯汀分校);VinAI Research(VinAI研究院)

Khai Nguyen, Dang Nguyen, Nhat Ho

AI总结:

针对最大切片 Wasserstein 距离的度量性缺失和摊销模型非对称问题,提出基于自注意力架构的 v-DSW 摊销优化,实现线性复杂度并提升点云重建性能。

摘要:

最大切片 Wasserstein(Max-SW)距离被广泛认为是解决切片 Wasserstein(SW)距离判别性不足的方案。在具有多个独立概率测度对的应用程序中,摊销投影优化被用于预测给定两个输入测度的“最大”投影方向,而不是多次使用投影梯度上升。尽管高效,Max-SW 及其摊销版本由于投影梯度上升的次优性和摊销差距而无法保证度量性质。因此,我们提出用带有 von Mises-Fisher(vMF)投影分布的分布切片 Wasserstein 距离(v-DSW)替代 Max-SW。由于 v-DSW 是任何非退化 vMF 分布的度量,其摊销版本在执行摊销时可以保证度量性。此外,当前的摊销模型不是置换不变和对称的。为了解决这个问题,我们基于自注意力架构设计了摊销模型。特别是,我们采用高效的自注意力架构,使计算在支撑点数量上呈线性。通过这两项改进,我们推导出自注意力摊销分布投影优化,并展示了其在点云重建及其下游应用中的优异性能。

英文摘要:

Max sliced Wasserstein (Max-SW) distance has been widely known as a solution for less discriminative projections of sliced Wasserstein (SW) distance. In applications that have various independent pairs of probability measures, amortized projection optimization is utilized to predict the ``max" projecting directions given two input measures instead of using projected gradient ascent multiple times. Despite being efficient, Max-SW and its amortized version cannot guarantee metricity property due to the sub-optimality of the projected gradient ascent and the amortization gap. Therefore, we propose to replace Max-SW with distributional sliced Wasserstein distance with von Mises-Fisher (vMF) projecting distribution (v-DSW). Since v-DSW is a metric with any non-degenerate vMF distribution, its amortized version can guarantee the metricity when performing amortization. Furthermore, current amortized models are not permutation invariant and symmetric. To address the issue, we design amortized models based on self-attention architecture. In particular, we adopt efficient self-attention architectures to make the computation linear in the number of supports. With the two improvements, we derive self-attention amortized distributional projection optimization and show its appealing performance in point-cloud reconstruction and its downstream applications.

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。

↑