(至此,GCN的回顾结束) 在关系抽取任务中,将一个句子构建的依赖树解析成一个图结构,该句子中的每一个token代表了图中的一个节点。如果单词之间具有依赖关系,那么在其对应的两个节点之间就存在一条边。在每一次图卷积操作之后,通过融合邻域节点的特征来更新每个节点的信息。 (下面介绍加权图卷积网络WGCN,Weighted gr...
Li等人[12]表明,GCN带来了许多卷积层过平滑的潜在问题。 “first-order” means the neighbor nodes to which the target node only need 1 step, and ”k-order” requires steps within distance k. 为了解决这些问题,我们提出了一种新的加权图卷积网络模型weighted graph convolutional network model(WGCN)。
论文地址 : http://ecai2020.eu/papers/957_paper.pdf论文源码 : https://github.com/LILI-ZHOU/EA-WGCN句子的形式化表示为 \chi = [x_1, x_2, \;...\; , x_i, \; ...\; , x_N ] , 其中 x_i 表示句子中第 i 个词, …
However, Most GCN-based methods use the fixed graph topology. Besides, only a single-scale feature is used, and the multi-scale information is ignored. In this paper, we propose a multi-scale skeleton adaptive weighted graph convolution network (MSAWGCN) for skeleton-based action recognition. ...
However, Most GCN-based methods use the fixed graph topology. Besides, only a single-scale feature is used, and the multi-scale information is ignored. In this paper, we propose a multi-scale skeleton adaptive weighted graph convolution network (MS-AWGCN) for skeleton-based action recognition....