In some image classification tasks, similarities among different categories are different and the samples are usually misclassified as highly similar categories. To distinguish highly similar categories, more specific features are required so that the classifier can improve the classification performance. In this paper, we propose a novel two-level hierarchical feature learning framework based on the deep convolutional neural network(CNN), which is simple and effective. First, the deep feature extractors of different levels are trained using the transfer learning method that fine-tunes the pre-trained deep CNN model toward the new target dataset. Second, the general feature extracted from all the categories and the specific feature extracted from highly similar categories are fused into a feature vector. Then the final feature representation is fed into a linear classifier. Finally, experiments using the Caltech-256, Oxford Flower-102, and Tasmania Coral Point Count(CPC) datasets demonstrate that the expression ability of the deep features resulting from two-level hierarchical feature learning is powerful. Our proposed method effectively increases the classification accuracy in comparison with flat multiple classification methods.
Network modeling is an important approach in many fields in analyzing complex systems. Recently new series of methods have emerged, by using Kronecker product and similar tools to model real systems. One of such approaches is the multiplicative attribute graph(MAG) model, which generates networks based on category attributes of nodes. In this paper we try to extend this model into a continuous one, give an overview of its properties, and discuss some special cases related to real-world networks, as well as the influence of attribute distribution and affinity function respectively.
The frequent outbreak of severe foodborne diseases(e.g., haemolytic uraemic syndrome and Listeriosis) in 2011 warns of a potential threat that world trade could spread fatal pathogens(e.g., enterohemorrhagic Escherichia coli). The epidemic potential from trade involves both intra-proliferation and inter-diffusion. Here, we present a worldwide vegetable trade network and a stochastic computational model to simulate global trade-mediated epidemics by considering the weighted nodes and edges of the network and the dual-scale dynamics of epidemics. We address two basic issues of network structural impact in global epidemic patterns:(1) in contrast to the prediction of heterogeneous network models, the broad variability of node degree and edge weights of the vegetable trade network do not determine the threshold of global epidemics;(2) a ‘penetration effect', by which community structures do not restrict propagation at the global scale, quickly facilitates bridging the edges between communities, and leads to synchronized diffusion throughout the entire network. We have also defined an appropriate metric that combines dual-scale behavior and enables quantification of the critical role of bridging edges in disease diffusion from widespread trading. The unusual structure mechanisms of the trade network model may be useful in producing strategies for adaptive immunity and reducing international trade frictions.
