Attributed graph clustering is challenging as it requires joint modelling of graph structures and node attributes. Recent progress on graph convolutional networks has proved that graph convolution is effective in combining structural and content information, and several recent methods based on it have achieved promising clustering performance on some real attributed networks. However, there is limited understanding of how graph convolution affects clustering performance and how to properly use it to optimize performance for different graphs. Existing methods essentially use graph convolution of a fixed and low order that only takes into account neighbours within a few hops of each node, which underutilizes node relations and ignores the diversity of graphs. In this paper, we propose an adaptive graph convolution method for attributed graph clustering that exploits high-order graph convolution to capture global cluster structure and adaptively selects the appropriate order for different graphs. We establish the validity of our method by theoretical analysis and extensive experiments on benchmark datasets. Empirical results show that our method compares favourably with state-of-the-art methods.

Graph-based methods have been demonstrated as one of the most effective approaches for semi-supervised learning, as they can exploit the connectivity patterns between labeled and unlabeled data samples to improve learning performance. However, existing graph-based methods either are limited in their ability to jointly model graph structures and data features, such as the classical label propagation methods, or require a considerable amount of labeled data for training and validation due to high model complexity, such as the recent neural-network-based methods. In this paper, we address label efficient semi-supervised learning from a graph filtering perspective. Specifically, we propose a graph filtering framework that injects graph similarity into data features by taking them as signals on the graph and applying a low-pass graph filter to extract useful data representations for classification, where label efficiency can be achieved by conveniently adjusting the strength of the graph filter. Interestingly, this framework unifies two seemingly very different methods -label propagation and graph convolutional networks. Revisiting them under the graph filtering framework leads to new insights that improve their modeling capabilities and reduce model complexity. Experiments on various semi-supervised classification tasks on four citation networks and one knowledge graph and one semi-supervised regression task for zero-shot image recognition validate our findings and proposals.

Many interesting problems in machine learning are being revisited with new deep learning tools. For graph-based semisupervised learning, a recent important development is graph convolutional networks (GCNs), which nicely integrate local vertex features and graph topology in the convolutional layers. Although the GCN model compares favorably with other state-of-the-art methods, its mechanisms are not clear and it still requires considerable amount of labeled data for validation and model selection. In this paper, we develop deeper insights into the GCN model and address its fundamental limits. First, we show that the graph convolution of the GCN model is actually a special form of Laplacian smoothing, which is the key reason why GCNs work, but it also brings potential concerns of oversmoothing with many convolutional layers. Second, to overcome the limits of the GCN model with shallow architectures, we propose both co-training and self-training approaches to train GCNs. Our approaches significantly improve GCNs in learning with very few labels, and exempt them from requiring additional labels for validation. Extensive experiments on benchmarks have verified our theory and proposals.

Medical visual question answering (Med-VQA) aims to accurately answer a clinical question presented with a medical image. Despite its enormous potential in healthcare industry and services, the technology is still in its infancy and is far from practical use. Med-VQA tasks are highly challenging due to the massive diversity of clinical questions and the disparity of required visual reasoning skills for different types of questions. In this paper, we propose a novel conditional reasoning framework for Med-VQA, aiming to automatically learn effective reasoning skills for various Med-VQA tasks. Particularly, we develop a question-conditioned reasoning module to guide the importance selection over multimodal fusion features. Considering the different nature of closed-ended and open-ended Med-VQA tasks, we further propose a type-conditioned reasoning module to learn a different set of reasoning skills for the two types of tasks separately. Our conditional reasoning framework can be easily applied to existing Med-VQA systems to bring performance gains. In the experiments, we build our system on top of a recent state-of-the-art Med-VQA model and evaluate it on the VQA-RAD benchmark [23]. Remarkably, our system achieves significantly increased accuracy in predicting answers to both closed-ended and open-ended questions, especially for open-ended questions, where a 10.8% increase in absolute accuracy is obtained. The source code can be downloaded from https://github.com/awenbocc/med-vqa. CCS CONCEPTS • Computing methodologies → Computer vision tasks.

The autoantibodies against C1q (anti-C1q) have been reported in patients with systemic lupus erythematosus (SLE). In the past decade, though there were increasing studies suggesting it is relatively specific in lupus nephritis (LN), its overall diagnostic value in LN has not been evaluated. The meta-analysis was conducted to quantitatively evaluate the diagnostic accuracy of autoantibodies against C1q in patients with LN, and to provide more precise evidence of a correlation between anti-C1q antibodies and activity of LN. We searched Medline, Embase and Cochrane databases and contacted authors if necessary. A total of 25 studies including 2,502 patients with SLE and 1,317 with LN met our inclusion criteria for this meta-analysis. Among all 25 studies, 22 studies were available for comparison between SLE with and without LN, and 9 studies compared anti-C1q between patients with active and inactive LN. Summary receiver operating characteristic (SROC) curve was used to summarize comprehensive test performance. The QUADAS tool was used to assess the quality of the studies. For the diagnosis of LN, the pooled sensitivity and specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), and diagnostic odds ratio (DOR) of anti-C1q were 0.58 (0.56-0.61, 95% confidence interval [95% CI]), 0.75 (0.72-0.77, 95% CI), 2.60 (2.06-3.28, 95% CI), 0.51 (0.41-0.63, 95% CI), and 6.08 (3.91-9.47, 95% CI) respectively. The area under the SROC curve (AUC) was 0.7941. For comparison between active and inactive LN, the weighted sensitivity, specificity, PLR, NLR and DOR were 0.74 (0.68-0.79, 95% CI), 0.77 (0.71-0.82, 95% CI), 2.91 (1.83-4.65, 95% CI), 0.33 (0.19-0.56, 95% CI), and 10.56 (4.56-24.46, 95% CI) respectively. The AUC was 0.8378. In conclusion, this meta-analysis indicates that anti-C1q antibodies have relatively fair sensitivity and specificity in the diagnosis of LN, suggesting that the presence of anti-C1q antibodies may be a valuable adjunct for predicting LN and assessing renal activity.