好消息 | 美图数据技术团队论文被AAAI-19录用

感谢阅读「美图数据技术团队」的第 26 篇原创文章，关注我们持续获取美图最新数据技术动态。

AAAI-19（Thirty-Third AAAI Conference on Artificial Intelligence）将于 2019 年 1 月 27日在美国夏威夷开幕，美图数据技术团队（MTdata）与中山大学 Inplus 实验室的合作论文「Tensor Decomposition for Multilayer Networks Clustering」被该会议录用。

AAAI 是人工智能领域的顶级国际学术会议，每年一届。据悉，今年 AAAI 共收到 7700 篇投稿，仅录用 1150 篇，录取率 16.2%创历史新低。

以下是该论文的简介：

Tensor Decomposition for Multilayer Networks Clustering

Zitai Chen, Chuan Chen, Zibin Zheng, Yi Zhu

多层网络上的聚类已被证实是一种能有效提高准确率的方法。各种多层网络聚类算法都是假设所有网络都源于潜在的聚类结构，并共同学习来自不同网络的兼容互补信息，以发掘一个共享的底层结构。然而，由于无关数据的存在，这个假设与许多现实中的实际应用相冲突。其中一个关键挑战是如何自动集成不同的数据表示来获得更好的预测表现。为此，我们提出了一种基于中心的多层网络聚类（CMNC）方法，它可以将不相关的关系划分为不同的网络组的同时揭示每个组中的聚类结构。多层网络在统一张量框架内表示，用于同时捕捉一组实体之间的多种类型的关系。通过使用非负约束的 rank-(Lr, Lr, 1)的张量分解（block term decomposition），我们能够基于图切割理论来很好的解释多个聚类结果。在数值上，我们将这个张量分解问题转化为了无约束的最优化问题，从而可以通过 nonlinear least squares (NLS)来求解。在大量的人工和真实数据集上的实验结果证实了我们的方法在面对无关数据时的有效性和鲁棒性。

附摘要原文： Clustering on multilayer networks has been shown to be a promising approach to enhance the accuracy. Various multilayer networks clustering algorithms assume all networks derive from a latent clustering structure, and jointly learn the compatible and complementary information from different networks to excavate one shared underlying structure. However, such assumptions is in conflict with many emerging reallife applications due to the existence of noisy/irrelevant networks. A key challenge here is to integrate different data representations automatically to achieve better predictive performance. To address this issue, we propose Centroidbased Multilayer Network Clustering (CMNC), a novel approach which can divide irrelevant relationships into different network groups and uncover the cluster structure in each group simultaneously. The multilayer networks is represented within a unified tensor framework for simultaneously capturing multiple types of relationships between a set of entities. By imposing the rank-(Lr , Lr , 1) block term decomposition with nonnegativity constraints, we are able to have well interpretations on the multiple clustering results based on graph cut theory. Numerically, we transform this tensor decomposition problem to an unconstrained optimization, thus can solve it efficiently under the nonlinear least squares (NLS) framework. Extensive experimental results on synthetic and realworld datasets show the effectiveness and robustness of our method against noise and irrelevant data.