VolterraNet: A Higher Order Convolutional Network With Group Equivariance for Homogeneous Manifolds

VolterraNet: A Higher Order Convolutional Network With Group Equivariance for Homogeneous Manifolds
复制标题

VolterraNet:具有同质流形群等方差的高阶卷积网络

DOI:
10.1109/tpami.2020.3035130
复制
发表时间:
2022
影响因子:
23.6
通讯作者:
Vemuri, Baba C.
Vemuri, Baba C.
中科院分区:
计算机科学1区
文献类型:
--
作者:
Banerjee, Monami;Chakraborty, Rudrasis;Bouza, Jose;Vemuri, Baba C.

文献摘要

参考文献

被引文献

相似文献

卷积神经网络由于其平移等变性,在基于图像的学习任务中取得了很大的成功。最近的工作将卷积神经网络的传统卷积层推广到非欧几里德空间,并证明了广义卷积运算的群等变性。针对黎曼齐次空间上定义为函数样本的数据,提出了一种新的高阶Volterra卷积神经网络(VolterraNet)。类似于传统卷积的结果,我们证明了Volterra泛函卷积与黎曼齐次空间所允许的等距群的作用是等变的,并且在一定的限制条件下,任何非线性等变函数都可以表示为我们的齐次空间Volterra卷积,推广了欧氏空间中Volterra展开式的非线性位移等变刻画。我们还证明了二阶泛函卷积运算可以表示为级联卷积运算,从而导致了一种有效的实现。此外,我们还提出了一个扩展的VolterraNet模型。这些进展导致了相对于基线非欧几里德CNN的大的参数减少。为了证明VolterraNet性能的有效性,我们提供了几个真实的数据实验,涉及球形MNIST、原子能、Shrec17数据集的分类任务,以及对扩散磁共振数据的分组测试。还给出了与最先进的性能比较。
Convolutional neural networks have been highly successful in image-based learning tasks due to their translation equivariance property. Recent work has generalized the traditional convolutional layer of a convolutional neural network to non-euclidean spaces and shown group equivariance of the generalized convolution operation. In this paper, we present a novel higher order Volterra convolutional neural network (VolterraNet) for data defined as samples of functions on Riemannian homogeneous spaces. Analagous to the result for traditional convolutions, we prove that the Volterra functional convolutions are equivariant to the action of the isometry group admitted by the Riemannian homogeneous spaces, and under some restrictions, any non-linear equivariant function can be expressed as our homogeneous space Volterra convolution, generalizing the non-linear shift equivariant characterization of Volterra expansions in euclidean space. We also prove that second order functional convolution operations can be represented as cascaded convolutions which leads to an efficient implementation. Beyond this, we also propose a dilated VolterraNet model. These advances lead to large parameter reductions relative to baseline non-euclidean CNNs. To demonstrate the efficacy of the VolterraNet performance, we present several real data experiments involving classification tasks on spherical-MNIST, atomic energy, Shrec17 data sets, and group testing on diffusion MRI data. Performance comparisons to the state-of-the-art are also presented.
测量 ARMA 模型组之间的距离
DOI: --
发表时间: 2016
期刊:
影响因子: --
作者:
U. Triacca
通讯作者: U. Triacca
DOI: --
发表时间: 2018-02
期刊: --
影响因子: --
作者:
Shaojie Bai;J. Z. Kolter;V. Koltun
通讯作者: Shaojie Bai;J. Z. Kolter;V. Koltun
DOI: --
发表时间: 1963
期刊:
影响因子: --
作者:
E. Hewitt;K. Ross
通讯作者: K. Ross
神经网络的 Volterra 表征
DOI: --
发表时间: 1991
期刊: [1991] Conference Record of the Twenty-Fifth Asilomar Conference on Signals, Systems & Computers
影响因子: --
作者:
N. Hakim;J. J. Kaufman;G. Cerf;H. Meadows
通讯作者: H. Meadows
使用递归超正方细分对 SO(3) 进行离散化
DOI: --
发表时间: 2017
期刊: International Conference on Multisensor Fusion and Integration for Intelligent Systems
影响因子: --
作者:
G. Kurz;F. Pfaff;U. Hanebeck
通讯作者: U. Hanebeck