论文阅读 [TPAMI-2022] Efficient and Stable Graph Scattering Transforms via Pruning

论文阅读 [TPAMI-2022] Efficient and Stable Graph Scattering Transforms via Pruning

论文搜索(studyai.com)

搜索论文: Efficient and Stable Graph Scattering Transforms via Pruning

搜索论文: http://www.studyai.com/search/whole-site/?q=Efficient+and+Stable+Graph+Scattering+Transforms+via+Pruning

关键字(Keywords)

Three-dimensional displays; Scattering; Transforms; Feature extraction; Stability analysis; Perturbation methods; Convolution

机器学习; 机器视觉

图网络; 图卷积网络; 三维点云; 点云分类

摘要(Abstract)

Graph convolutional networks (GCNs) have well-documented performance in various graph learning tasks, but their analysis is still at its infancy.

图卷积网络（GCN）在各种图学习任务中有着良好的表现，但它们的分析仍处于起步阶段。.

Graph scattering transforms (GSTs) offer training-free deep GCN models that extract features from graph data, and are amenable to generalization and stability analyses.

图散射变换（GST）提供了无需训练的深层GCN模型，可以从图数据中提取特征，并且易于泛化和稳定性分析。.

The price paid by GSTs is exponential complexity in space and time that increases with the number of layers.

GST付出的代价是空间和时间的指数复杂性，随着层数的增加而增加。.

This discourages deployment of GSTs when a deep architecture is needed.

当需要深层架构时，这会阻碍GST的部署。.

The present work addresses the complexity limitation of GSTs by introducing an efficient so-termed pruned §GST approach.

目前的工作通过引入一种高效的所谓修剪（p）GST方法来解决GST的复杂性限制。.

The resultant pruning algorithm is guided by a graph-spectrum-inspired criterion, and retains informative scattering features on-the-fly while bypassing the exponential complexity associated with GSTs.

由此产生的剪枝算法受图形频谱启发准则的指导，并在飞行中保留信息性散射特征，同时绕过与GST相关的指数复杂性。.

Stability of the novel pGSTs is also established when the input graph data or the network structure are perturbed.

当输入图形数据或网络结构受到扰动时，新pGSTs的稳定性也得以建立。.

Furthermore, the sensitivity of pGST to random and localized signal perturbations is investigated analytically and experimentally.

此外，还对pGST对随机和局部信号扰动的灵敏度进行了分析和实验研究。.

Numerical tests showcase that pGST performs comparably to the baseline GST at considerable computational savings.

数值试验表明，pGST在相当大的计算节省下与基线GST的性能相当。.

Furthermore, pGST achieves comparable performance to state-of-the-art GCNs in graph and 3D point cloud classification tasks.

此外，pGST在图形和3D点云分类任务中的性能与最先进的GCN相当。.

Upon analyzing the pGST pruning patterns, it is shown that graph data in different domains call for different network architectures, and that the pruning algorithm may be employed to guide the design choices for contemporary GCNs…

通过分析pGST剪枝模式，发现不同领域的图形数据需要不同的网络结构，剪枝算法可以用来指导当代GCN的设计选择。。.

作者(Authors)

[‘Vassilis N. Ioannidis’, ‘Siheng Chen’, ‘Georgios B. Giannakis’]