The problem of extracting meaningful data through graph analysis spans a range of different fields, such as the internet, social networks, biological networks, and many others. The importance of ...being able to effectively mine and learn from such data continues to grow as more and more structured data become available. In this paper, we present a simple and scalable semi-supervised learning method for graph-structured data in which only a very small portion of the training data are labeled. To sufficiently embed the graph knowledge, our method performs graph convolution from different views of the raw data. In particular, a dual graph convolutional neural network method is devised to jointly consider the two essential assumptions of semi-supervised learning: (1) local consistency and (2) global consistency. Accordingly, two convolutional neural networks are devised to embed the local-consistency-based and global-consistency-based knowledge, respectively. Given the different data transformations from the two networks, we then introduce an unsupervised temporal loss function for the ensemble. In experiments using both unsupervised and supervised loss functions, our method outperforms state-of-the-art techniques on different datasets.
A travel route recommendation service that recommends a sequence of points of interest for tourists traveling in an unfamiliar city is a very useful tool in the field of location-based social ...networks. Although there are many web services and mobile applications that can help tourists to plan their trips by providing information about sightseeing attractions, travel route recommendation services are still not widely applied. One reason could be that most of the previous studies that addressed this task were based on the orienteering problem model, which mainly focuses on the estimation of a user-location relation (for example, a user preference). This assumes that a user receives a reward by visiting a point of interest and the travel route is recommended by maximizing the total rewards from visiting those locations. However, a location-location relation, which we introduce as a transition pattern in this paper, implies useful information such as visiting order and can help to improve the quality of travel route recommendations. To this end, we propose a travel route recommendation method by combining location and transition knowledge, which assigns rewards for both locations and transitions.
In this paper, we propose a robust visual object clustering approach based on bounding box ranking to discover the characteristics of objects from real-world datasets containing a large number of ...noisy images, and apply it to sightseeing spot assessment. The purpose is to develop a diversity of resources for sightseeing from images available on social network services (SNS). Objects appearing frequently in images captured in a certain city may represent a certain characteristic of it (local culture, architecture, and so on). Such knowledge can be used to discover various sightseeing resources from the perspective of the user rather than that of the provider (e.g., a travel agency). However, owing to the variable quality of images on SNS, it is challenging to identify objects common to several images by using conventional object discovery methods, and this is where the proposed approach is useful. Extensive experiments on standard and extended benchmarks verified its effectiveness. We also tested the proposed method on an application where the characteristics of a city (i.e., cultural elements) were discovered from a set of images of it. Moreover, by utilizing the objects discovered from images on SNS, we propose an object-level assessment framework to rank sightseeing spots by assigning scores and verify its performance.
Graph embedding aims at learning representations of nodes in a low dimensional vector space. Good embeddings should preserve the graph topological structure. To study how much such structure can be ...preserved, we propose evaluation methods from four aspects: 1) How well the graph can be reconstructed based on the embeddings, 2) The divergence of the original link distribution and the embedding-derived distribution, 3) The consistency of communities discovered from the graph and embeddings, and 4) To what extent we can employ embeddings to facilitate link prediction. We find that it is insufficient to rely on the embeddings to reconstruct the original graph, to discover communities, and to predict links at a high precision. Thus, the embeddings by the state-of-the-art approaches can only preserve part of the topological structure.
nema
Actionable knowledge graph (AKG), a specialized version of knowledge graph, was proposed recently to represent, analyze, and predict human action, thus facilitating deeper understanding of human ...action by robots. However, the automatic construction of AKGs from action-related corpora is still an unexplored problem. In this study, we first propose three unsupervised matrix factorization-based frameworks for AKG generation from three different perspectives: subject, context and functionality of action, respectively. Further, we propose a hybrid model based on neural network matrix factorization (NNMF) that considers multi-source signals simultaneously. It not only learns the latent action representations, but also learns the optimal learning objective rather than assuming it to be fixed. To quantitatively verify the utility of the constructed AKGs, we introduce a novel application, that is, predicting the most likely missing action records in Wikipedia biographies. Experimental results on a large-scale Wikipedia biography dataset show that the proposed model brings significant improvement over the baselines, which demonstrates the strong expressiveness of our generated AKGs.
Graph neural networks (GNNs) have recently made remarkable breakthroughs in the paradigm of learning with graph-structured data. However, most existing GNNs limit the receptive field of the node on ...each layer to its connected (one-hop) neighbors, which disregards the fact that large receptive field has been proven to be a critical factor in state-of-the-art neural networks. In this paper, we propose a novel approach to appropriately define a variable receptive field for GNNs by incorporating high-order proximity information extracted from the hierarchical topological structure of the input graph. Specifically, multiscale groups obtained from trainable hierarchical semi-nonnegative matrix factorization are used for adjusting the weights when aggregating one-hop neighbors. Integrated with the graph attention mechanism on attributes of neighboring nodes, the learnable parameters within the process of aggregation are optimized in an end-to-end manner. Extensive experiments show that the proposed method (hpGAT) outperforms state-of-the-art methods and demonstrate the importance of exploiting high-order proximity in handling noisy information of local neighborhood.
Neural network pruning plays an important role in the deployment on resource-constrained devices by reducing the scale of the network and the computational complexity. Different from existing pruning ...methods that only consider the amount of information filters contain, we think the distribution of information is more important to the model accuracy. Then we use visualization of feature map to consider the impact of the attention area on the network decision. In order to directly relate the attention area with confidence score, we propose HScore to use interpolation and normalization to visualize the attention area of feature map and calculate the confidence score by putting only the attention area into the model. The principle behind is that if the feature map focuses on an area which can achieve a higher classification confidence, then the corresponding channel is important to keep the accuracy of model. The confidence score of attention area represents the importance of the channel to the model and filters with smaller confidences would be pruned. The experiment results prove that HScore can prune more redundant filters while maintaining better accuracy. Notably, our method can reduce 68.6% parameters and 73.0% FLOPs of ResNet-110 even with 0.14% top-1 accuracy increase on CIFAR-10. With ResNet-50, we achieve a 55% FLOPs reduction with only a loss of 1.03% in the top-1 accuracy on LSVRC-2012, which has advanced the state-of-the-art.
Full text
Available for:
GEOZS, IJS, IMTLJ, KILJ, KISLJ, NLZOH, NUK, OILJ, PNG, SAZU, SBCE, SBJE, UILJ, UL, UM, UPCLJ, UPUK, ZAGLJ, ZRSKP
Vortex-Induced Vibration (VIV) caused by ocean current is one of the main sources of the long-term fatigue damage of marine riser. Buoyancy module is an essential equipment for deep-water riser ...system, which can reduce the top tension remand. Based on the hydrodynamic force decomposition model and forced vibration test data, a time domain prediction model of VIV response for top-tensioned riser with buoyancy modules is established in this paper. Firstly, the numerical model is verified with the experimental measurements of an 11.48 m riser with buoyancy modules under uniform flow. Then, a 610 m riser is chosen as the research object, and five staggered buoyancy module configurations with different coverage ratios are simulated, to reveal the VIV response characteristics of full-scale riser with buoyancy modules. Some interesting phenomena are reasonably explained with VIV mechanism analysis. In terms of frequency response, displacement and modal order, the influence of coverage ratio on riser's VIV is analyzed. Finally, the optimal coverage of buoyancy modules for VIV suppression is discussed, which is of great significance in engineering application.
•A time domain prediction model is proposed to investigate VIV response of top-tensioned riser with buoyancy modules.•VIV response characteristics of a full-scale riser with buoyancy modules are analyzed systematically.•The buoyancy coverage ratio effect on riser's VIV characteristics is analyzed.•Some interesting phenomena are reasonably explained with VIV mechanism.•The optimal buoyancy modules coverage for VIV suppression is discussed.
Full text
Available for:
GEOZS, IJS, IMTLJ, KILJ, KISLJ, NLZOH, NUK, OILJ, PNG, SAZU, SBCE, SBJE, UILJ, UL, UM, UPCLJ, UPUK, ZAGLJ, ZRSKP
We propose a general view that demonstrates the relationship between network embedding approaches and matrix factorization. Unlike previous works that present the equivalence for the approaches from ...a skip-gram model perspective, we provide a more fundamental connection from an optimization (objective function) perspective. We demonstrate that matrix factorization is equivalent to optimizing two objectives: one is for bringing together the embeddings of similar nodes; the other is for separating the embeddings of distant nodes. The matrix to be factorized has a general form: S-β. The elements of $\mathbfS $ indicate pairwise node similarities. They can be based on any user-defined similarity/distance measure or learned from random walks on networks. The shift number β is related to a parameter that balances the two objectives. More importantly, the resulting embeddings are sensitive to β and we can improve the embeddings by tuning β. Experiments show that matrix factorization based on a new proposed similarity measure and β-tuning strategy significantly outperforms existing matrix factorization approaches on a range of benchmark networks.
PDF
