arXiv subjects

Tsuyoshi Murata

Publications and source records attributed to Tsuyoshi Murata.

26 records · Page 2Linked to original sources

arXiv2020

MetAL: Active Semi-Supervised Learning on Graphs via Meta Learning

The objective of active learning (AL) is to train classification models with less number of labeled instances by selecting only the most informative instances for labeling. The AL algorithms designed for other data types such as images and text do not perform well on graph-structured data. Although a few heuristics-based AL algorithms have been proposed for graphs, a principled approach is lacking. In this paper, we propose MetAL, an AL approach that selects unlabeled instances that directly improve the future performance of a classification model. For a semi-supervised learning problem, we formulate the AL task as a bilevel optimization problem. Based on recent work in meta-learning, we use the meta-gradients to approximate the impact of retraining the model with any unlabeled instance on the model performance. Using multiple graph datasets belonging to different domains, we demonstrate that MetAL efficiently outperforms existing state-of-the-art AL algorithms.

stat.ML↗

arXiv2019

Learning Graph Neural Networks with Noisy Labels

We study the robustness to symmetric label noise of GNNs training procedures. By combining the nonlinear neural message-passing models (e.g. Graph Isomorphism Networks, GraphSAGE, etc.) with loss correction methods, we present a noise-tolerant approach for the graph classification task. Our experiments show that test accuracy can be improved under the artificial symmetric noisy setting.

cs.LG↗

arXiv2018

Exploring Partially Observed Networks with Nonparametric Bandits

Real-world networks such as social and communication networks are too large to be observed entirely. Such networks are often partially observed such that network size, network topology, and nodes of the original network are unknown. In this paper we formalize the Adaptive Graph Exploring problem. We assume that we are given an incomplete snapshot of a large network and additional nodes can be discovered by querying nodes in the currently observed network. The goal of this problem is to maximize the number of observed nodes within a given query budget. Querying which set of nodes maximizes the size of the observed network? We formulate this problem as an exploration-exploitation problem and propose a novel nonparametric multi-arm bandit (MAB) algorithm for identifying which nodes to be queried. Our contributions include: (1) $i$KNN-UCB, a novel nonparametric MAB algorithm, applies $k$-nearest neighbor UCB to the setting when the arms are presented in a vector space, (2) provide theoretical guarantee that $i$KNN-UCB algorithm has sublinear regret, and (3) applying $i$KNN-UCB algorithm on synthetic networks and real-world networks from different domains, we show that our method discovers up to 40% more nodes compared to existing baselines.

stat.ML↗

arXiv2014

A framework for community detection in heterogeneous multi-relational networks

There has been a surge of interest in community detection in homogeneous single-relational networks which contain only one type of nodes and edges. However, many real-world systems are naturally described as heterogeneous multi-relational networks which contain multiple types of nodes and edges. In this paper, we propose a new method for detecting communities in such networks. Our method is based on optimizing the composite modularity, which is a new modularity proposed for evaluating partitions of a heterogeneous multi-relational network into communities. Our method is parameter-free, scalable, and suitable for various networks with general structure. We demonstrate that it outperforms the state-of-the-art techniques in detecting pre-planted communities in synthetic networks. Applied to a real-world Digg network, it successfully detects meaningful communities.

cs.SI↗

arXiv2014

Detecting network communities beyond assortativity-related attributes

In network science, assortativity refers to the tendency of links to exist between nodes with similar attributes. In social networks, for example, links tend to exist between individuals of similar age, nationality, location, race, income, educational level, religious belief, and language. Thus, various attributes jointly affect the network topology. An interesting problem is to detect community structure beyond some specific assortativity-related attributes $ρ$, i.e., to take out the effect of $ρ$ on network topology and reveal the hidden community structure which are due to other attributes. An approach to this problem is to redefine the null model of the modularity measure, so as to simulate the effect of $ρ$ on network topology. However, a challenge is that we do not know to what extent the network topology is affected by $ρ$ and by other attributes. In this paper, we propose Dist-Modularity which allows us to freely choose any suitable function to simulate the effect of $ρ$. Such freedom can help us probe the effect of $ρ$ and detect the hidden communities which are due to other attributes. We test the effectiveness of Dist-Modularity on synthetic benchmarks and two real-world networks.

cs.SI↗

arXiv2013

Extending modularity by capturing the similarity attraction feature in the null model

Modularity is a widely used measure for evaluating community structure in networks. The definition of modularity involves a comparison of within-community edges in the observed network and that number in an equivalent randomized network. This equivalent randomized network is called the null model, which serves as a reference. To make the comparison significant, the null model should characterize some features of the observed network. However, the null model in the original definition of modularity is unrealistically mixed, in the sense that any node can be linked to any other node without preference and only connectivity matters. Thus, it fails to be a good representation of real-world networks. A common feature of many real-world networks is "similarity attraction", i.e., edges tend to link to nodes that are similar to each other. We propose a null model that captures the similarity attraction feature. This null model enables us to create a framework for defining a family of Dist-Modularity adapted to various networks, including networks with additional information on nodes. We demonstrate that Dist-Modularity is useful in identifying communities at different scales.

cs.SI↗

arXiv2010

Detecting Communities in Tripartite Hypergraphs

In social tagging systems, also known as folksonomies, users collaboratively manage tags to annotate resources. Naturally, social tagging systems can be modeled as a tripartite hypergraph, where there are three different types of nodes, namely users, resources and tags, and each hyperedge has three end nodes, connecting a user, a resource and a tag that the user employs to annotate the resource. Then, how can we automatically detect user, resource and tag communities from the tripartite hypergraph? In this paper, by turning the problem into a problem of finding an efficient compression of the hypergraph's structure, we propose a quality function for measuring the goodness of partitions of a tripartite hypergraph into communities. Later, we develop a fast community detection algorithm based on minimizing the quality function. We explain advantages of our method and validate it by comparing with various state of the art techniques in a set of synthetic datasets.

cs.SI↗

arXiv2010

Advanced modularity-specialized label propagation algorithm for detecting communities in networks

A modularity-specialized label propagation algorithm (LPAm) for detecting network communities was recently proposed. This promising algorithm offers some desirable qualities. However, LPAm favors community divisions where all communities are similar in total degree and thus it is prone to get stuck in poor local maxima in the modularity space. To escape local maxima, we employ a multistep greedy agglomerative algorithm (MSG) that can merge multiple pairs of communities at a time. Combining LPAm and MSG, we propose an advanced modularity-specialized label propagation algorithm (LPAm+). Experiments show that LPAm+ successfully detects communities with higher modularity values than ever reported in two commonly used real-world networks. Moreover, LPAm+ offers a fair compromise between accuracy and speed.

physics.soc-ph↗