Deep learning has revolutionized many machine learning tasks in recent years, ranging from image classification and video processing to speech recognition and natural language understanding. The data in these tasks are typically represented in the Euclidean space. However, there is an increasing number of applications where data are generated from non-Euclidean domains and are represented as graphs with complex relationships and interdependency between objects. The complexity of graph data has imposed significant challenges on existing machine learning algorithms. Recently, many studies on extending deep learning approaches for graph data have emerged. In this survey, we provide a comprehensive overview of graph neural networks (GNNs) in data mining and machine learning fields. We propose a new taxonomy to divide the state-of-the-art graph neural networks into four categories, namely recurrent graph neural networks, convolutional graph neural networks, graph autoencoders, and spatial-temporal graph neural networks. We further discuss the applications of graph neural networks across various domains and summarize the open source codes, benchmark data sets, and model evaluation of graph neural networks. Finally, we propose potential research directions in this rapidly growing field.

No abstract

Abstract:A large body of evidence demonstrates the key role played by entrepreneurship in promoting economic growth. However, the potential connections between entrepreneurship, social networking, and economic development still require in-depth exploration and discussion. This paper first establishes a theoretical framework combining entrepreneurship capital theory, resource dependence theory and transaction cost theory, then examines the possible associations between entrepreneurship, social networks, and economic growth based on the dynamic panel data model. To achieve the research objectives, the investigators collected data spanning the period between 2007 and 2016 from 31 provinces and cities in China. The authors adopted the enterprise employment rate as a measure of entrepreneurship and used the information sharing rate to assess social networks, which were then both introduced into the economic growth model. Additionally, by using the system of generalized method of moments (GMM) estimation, this article measures the influence of entrepreneurship and social networks on the economic growth of a local area. The empirical results reveal that both entrepreneurship and social networking significantly promote regional economic growth in China. Further, the effect of entrepreneurship is significantly enhanced after introducing the joint effects of entrepreneurship and social network. The findings also expound that entrepreneurship of the eastern zone and social networking of the central section exhibit the strongest potential for economic development of the respective areas. Conversely, entrepreneurship may actually hinder the economic advancement of the central areas of China. Corresponding to the findings, the researchers suggest that it is necessary to devise flexible policies for heterogeneous entrepreneurial environments and to appropriately utilize interpersonal networks to maximize the efficiency of the outputs of economic activity, which are likely to strengthen the role of entrepreneurship and social networks in contemporary economic and business milieu.

To investigate the impact of institutional investors on firms' corporate social responsibility (CSR) engagement while controlling for possible endogeneity concerns, we study how Chinese listed firms adjust their CSR decisions when their institutional investors are distracted by exogenous attention-grabbing events and thus are inattentive. With a sample of Chinese listed firms from 2009 to 2017, we find a significant and robust negative relationship between institutional investor inattention and firms' CSR engagement. This negative relationship is more pronounced for firms with more principal-agent problems and/or weaker corporate governances and is more attributable to the inattention of institutional investors with more monitoring incentives.These findings suggest that managers are less motivated to engage in CSR when they are less monitored by institutional investors, indicating that CSR is beneficial to shareholders of Chinese listed firms. Our findings also indicate that the positive impact of institutional investors on CSR may be constrained by their limited attention.

Non-performing loans of commercial banks have long hampered the development of the banking sector, and directly reflect the credit risk and asset quality. With the continuous development of the financial industry, the introduction of financial inclusion has greatly eased the shortage of funds, and narrowed the gap between poor and rich. However, whether the promotion of financial inclusion in the financial industry could affect the non-performing loans of commercial banks has not been verified. Therefore, this paper discusses the possible associations between financial inclusion and non-performing loans of commercial banks on the regional level, constructs a panel data model by selecting the data of 31 provinces (including 4 municipalities) in China from 2005 to 2016, and uses the fixed effect model for empirical test. The empirical results (from an overall national sample) reveal a negative impact of the financial inclusion on non-performing loans. Moreover, the development of the banking sector and the regional consumption could enhance the impact of financial inclusion, while government intervention and unemployment could reduce the impact of financial inclusion. From the analysis of the regional sample, when the development of financial inclusion reaches a high level, the lagged financial inclusion promote the non-performing loans of commercial banks; however, when the financial inclusion is underdeveloped, the development of commercial banks act as a disincentive to non-performing loans. Therefore, the local governments should pay more attention to the influences of financial inclusion on the financial industry, in order to maintain the stability of banking asset quality. In addition, the negative impact of financial inclusion on non-performing loans of commercial banks is significant in China central region, while its impacts in China eastern and western regions are not significant. This indicates that the development of the financial industry and economy can hamper the effects of financial inclusion. It is necessary to adjust the financial resource allocation according to the characteristics of different regions in China, so that the financial inclusion can effectively promote the regional financial industry upgrade, improve regional capital flow efficiency, and fundamentally reduce the non-performing loans of commercial banks. According to the sample analysis by time, there is a significant negative impact relationship between inclusive finance and commercial banks’ non-performing loans after the financial crisis, while the impacts before and during the financial crisis are not significant. This demonstrates that the impact of the global financial crisis on China’s regional economy has further enhanced the inefficiency of the inclusive financial system on credit risk, which in turn, helps commercial banks better maintain asset quality stability.

Graph convolutional networks (GCNs) have recently become one of the most powerful tools for graph analytics tasks in numerous applications, ranging from social networks and natural language processing to bioinformatics and chemoinformatics, thanks to their ability to capture the complex relationships between concepts. At present, the vast majority of GCNs use a neighborhood aggregation framework to learn a continuous and compact vector, then performing a pooling operation to generalize graph embedding for the classification task. These approaches have two disadvantages in the graph classification task: (1)when only the largest sub-graph structure (k-hop neighbor) is used for neighborhood aggregation, a large amount of early-stage information is lost during the graph convolution step; (2) simple average/sum pooling or max pooling utilized, which loses the characteristics of each node and the topology between nodes. In this paper, we propose a novel framework called, dual attention graph convolutional networks (DAGCN) to address these problems. DAGCN automatically learns the importance of neighbors at different hops using a novel attention graph convolution layer, and then employs a second attention component, a selfattention pooling layer, to generalize the graph representation from the various aspects of a matrix graph embedding. The dual attention network is trained in an end-to-end manner for the graph classification task. We compare our model with state-of-the-art graph kernels and other deep learning methods. The experimental results show that our framework not only outperforms other baselines but also achieves a better rate of convergence.

The upgrading of industrial structure is the core means of coordinating economic development and environment protection. Its spatial agglomeration can also reduce environmental pollution partly. The upgrading of China’s industrial structure has become an important issue concerned by the whole society. To better understand this issue, based on the provincial data of China (1997–2017), this paper strives to explore the spatial effects of foreign trade and foreign direct investment (FDI) on the upgrading of China’s regional industrial structure by constructing the weight matrix of economic distance, and by introducing the spatial autocorrelation analysis method and spatial panel econometric model. The results show that: 1. The Moran’s I index of China’s import, export, FDI, and industrial structure upgrading has passed the 5% significance level test, displaying remarkable spatial agglomeration characteristics. 2. Foreign trade and FDI are important driving factors to upgrade China’s industrial structure. 3. Foreign trade has a significant spatial spillover effect. Imports and exports can not only promote the upgrading of local industrial structure, but also radiate to other regions, promote or inhibit the development of its industry, and further affect the national data. 4. The spatial spillover effect of FDI is not significant. Finally, some policy suggestions are put forward.

High-polluting industries are important sources of pollutant emissions, and closely related to many environmental issues. High-polluting firms face the pressure to exploit technological innovation for improving their environmental operations. This paper explores the impact of corporate social responsibility and public attention on the innovation performance of high-polluting firms. Based on a sample of China’s listed firms in high-polluting industries from 2011 to 2016, we use a panel data model to investigate the associations among corporate social responsibility, public attention and innovation performance. The results show that there is a positive association between corporate social responsibility and innovation performance. There is a positive association between public attention and innovation performance as well. The pressure of regional economies can hinder innovation performance. Furthermore, in the subsample of state-owned enterprises, the association between public attention and innovation performance is more pronounced. Meanwhile, the corporate social responsibility of non-state-owned enterprises plays a stronger role for innovation performance, but its effect will be limited by the pressure of regional economies. Our results can help high-polluting firms implement the innovation strategies for obtaining more environmental benefits and achieving sustainable development.