Search

Visual urban perception has recently attracted a lot of research attention owing to its importance in many fields. Traditional methods for visual urban perception mostly need to collect adequate training instances for newly-added perception attributes. In this paper, we consider a novel formulation, zero-shot learning, to free this cumbersome curation. Based on the idea of different images containing similar objects are more likely to possess the same perceptual attribute, we learn the semantic correlation space formed by objects semantic information and perceptual attributes. For newly-added attributes, we attempt to synthesize their prototypes by transferring similar object vector representations between the unseen attributes and the training (seen) perceptual attributes. For this purpose, we leverage a deep semantic-aware network for zero-shot visual urban perception model. It is a new two step zero-shot learning architecture, which includes supervised visual urban perception step for training attributes and zero-shot prediction step for unseen attributes. In the first step, we highlight the important role of semantic information and introduce it into supervised deep visual urban perception framework for training attributes. In the second step, we use the visualization techniques to obtain the correlations between semantic information and visual perception attributes from the well trained supervised model, and learn the prototype of unseen attributes and testing images to predict perception score on unseen attributes. The experimental results on a large-scale benchmark dataset validate the effectiveness of our method.

The methylation of arginine residues by protein arginine methyltransferases (PRMTs) is a crucial post-translational modification for many biological processes, including DNA repair, RNA processing, and transduction of intra- and extracellular signaling. Previous studies have reported that PRMTs are extensively involved in various pathologic states, including cancer, inflammation, and oxidative stress reaction. However, the role of PRMTs has not been well described in kidney diseases. Recent studies have shown that aberrant function of PRMTs and its metabolic products—symmetric dimethylarginine (SDMA) and asymmetric dimethylarginine (ADMA)—are involved in several renal pathological processes, including renal fibrosis, acute kidney injury (AKI), diabetic nephropathy (DN), hypertension, graft rejection and renal tumors. We aim in this review to elucidate the possible roles of PRMTs in normal renal function and various kidney diseases.

People used to take a series of nearly identical photos about the same subject, but it is usually a tedious chore to select the reversed ones from them. Despite the remarkable progress, most existing studies on image aesthetics assessment fail to fulfill the task of series photo selection. In this paper, we develop a novel deep CNN architecture that aggregates multi-scale features from different network layers, in order to capture the subtle differences between series photos. To reduce the risk of redundant or even interfering features, we introduce the spatial-channel self-attention mechanism to adaptively recalibrate the features at each layer, so that informative features can be selectively emphasized and less useful ones suppressed. Extensive experiments on a benchmark dataset well demonstrate the potential of our approach for series photo selection.

Relation classification is a crucial ingredient in numerous information-extraction systems and has attracted a great deal of attention in recent years. Traditional approaches largely rely on feature engineering and suffer from the limitations of the domain adaption and the error propagation. To overcome the above-mentioned problems, many deep neural network-based methods have been proposed; however, these methods cannot effectively locate and utilize the relation trigger features. To locate the relation trigger features and make full use of them, we propose a novel multi-gram convolution neural network-based self-attention model with a recurrent neural network framework. The multi-gram conventional neural network attention model can learn the adaptive relational semantics of inputs based on the fact that a relation can be totally defined by the shortest dependency path between its two entities. With the learned relational semantics, we can obtain the corresponding importance distribution over input sentences and locate the relation trigger features. For effective information propagation and integration, we utilize a bidirectional gated recurrent unit to encode the high-level features during recurrent propagation. The experimental results on two benchmark datasets demonstrate that the proposed model outperforms most of the state-of-the-art models.

Robust texture analysis of multi-modal images is important for practical breast tumor diagnosis applications. Texture features based on ranklet transform were proposed for breast tumor classification of multi-modal images and improved diagnostic performance. However, two limitations still exist in these features. Ranklet transform ignores local characteristics of images which are important for texture feature extraction. In addition, due to application of multi-resolution analysis of ranklet transform, some noises or redundant information may be introduced. These issues may result in performance degradation. To solve these problems, this paper proposes a robust texture feature based on Local Structure Preserving Ranklet (LSP-Ranklet) transform and multi-task learning. First of all, multiple LSP-Ranklet images are generated via LSP-Ranklet transform. In this procedure, the distance-based weighting method is proposed to preserve local structure of images by learning local relevance between pixels. Based on LSP-Ranklet images, texture features based on Gray-Level Co-occurrence Matrix (GLCM) are extracted. To eliminate noises of extracted features, multi-task feature learning is employed to select common feature subsets which are robust for tumor classification of multi-modal images. At last, SVM model is used for tumor classification. Experimental results on our multi-modal breast ultrasound images database demonstrate the effectiveness and robustness of the proposed feature.

Large-scale knowledge graphs (KGs) are shown to become more important in current information systems. To expand the coverage of KGs, previous studies on knowledge graph completion need to collect adequate training instances for newly-added relations. In this paper, we consider a novel formulation, zero-shot learning, to free this cumbersome curation. For newly-added relations, we attempt to learn their semantic features from their text descriptions and hence recognize the facts of unseen relations with no examples being seen. For this purpose, we leverage Generative Adversarial Networks (GANs) to establish the connection between text and knowledge graph domain: The generator learns to generate the reasonable relation embeddings merely with noisy text descriptions. Under this setting, zero-shot learning is naturally converted to a traditional supervised classification task. Empirically, our method is model-agnostic that could be potentially applied to any version of KG embeddings, and consistently yields performance improvements on NELL and Wiki dataset.

Wilson’s disease (WD) is a genetic disorder of copper metabolism with pathological copper accumulation in the brain and any other tissues. This article aimed to assess lesions in cerebello-thalamo-cortical network with an advanced technique of diffusion tensor imaging (DTI) in WD. 35 WD patients and 30 age- and sex-matched healthy volunteers were recruited to accept diffusion-weighted images with 15 gradient vectors and conventional magnetic resonance imaging (MRI). The DTI parameters, including fractional anisotropy (FA) and mean diffusion (MD), were calculated by diffusion kurtosis estimator software. After registration, patient groups with FA mappings and MD mappings and normal groups were compared with 3dttest and receiver-operating characteristic (ROC) curve analysis, corrected with FDR simulations (p=0.001, α=0.05, cluster size = 326). We found that the degree of FA increased in the bilateral head of the caudate nucleus (HCN), lenticular nucleus (LN), ventral thalamus, substantia nigra (SN), red nucleus (RN), right dentate nucleus (DN), and decreased in the mediodorsal thalamus and extensive white matter. The value of MD increased in HCN, LN, SN, RN, and extensive white matter. The technique of DTI provides higher sensitivity and specificity than conventional MRI to detect Wilson’s disease. Besides, lesions in the basal ganglia, thalamus, and cerebellum might disconnect the basal ganglia-thalamo-cortical circuits or dentato-rubro-thalamic (DRT) track and disrupt cerebello-thalamo-cortical network finally, which may cause clinical extrapyramidal symptoms.

In the template matching of finger vein recognition, the probe will be accepted if the number of its overlapped vein points with the enrolled user is larger than the predefined threshold. However, the acceptance may be false owing to ignoring the structure of the vein pattern. We find that local vein branches near the bifurcation point of vein pattern vary largely between the imposter images. So, this paper tries to explore this kind of local vein structure to improve the recognition performance of the template matching. The bifurcation point and its local vein branches, named tri-branch vein structure, are extracted from the vein pattern, and fused with the whole vein pattern by a user-specific threshold-based filter framework. The experimental results on two public databases prove the effectiveness of the proposed framework for improving the performance of vein pattern-based finger vein recognition.

Student performance prediction is of great importance to many educational domains, such as academic early warning and personalized teaching, and has drawn numerous research attention in recent decades. Most of the previous studies are based on students’ historical course grades, demographical data, in-class study performance, and online activities from e-learning platforms, e.g., Massive Open Online Courses (MOOCs). Thanks to the widely used of campus smartcard, it supplies an opportunity to predict students’ academic performance with their off-line behavioral data. In this study, we seek to capture three student behavioral characters, including duration, variation and periodicity, and predict students’ performance based on the three types of information. However, it is highly challenging to extract efficient features manually from the huge amount of raw smartcard records. Besides, it is not trivial to construct a good predictive model for some majors with limited student samples. To address the above issues, we develop a novel end-to-end deep learning method and propose Dual Path Convolutional Neural Network (DPCNN) for student performance prediction. Moreover, we introduce multi-task learning to our method and predict the performance of students from different majors in a unified framework. Experimental results demonstrate the superiority of our approach over the state-of-the-art methods.

Overweight and obesity are dramatically increasing worldwide. In addition to being the most important factor for the increase in diabetes prevalence, there is a growing evidence of obesity being also significantly associated with the risks and poor outcome in ovarian cancer (OVC). Metformin is the most widely used first-line type 2 diabetes drug, currently being studied for its association with the decreased risk of occurrence and better survival of OVC patients. In this review, we discussed the proposed mechanisms of metformin-exerted anticancer effects, as well as the preclinical and clinical data suggesting its beneficial effect against this devastating condition.

A new trigger words identification method is defined based on word activation force.A unique representation of activation force defined dependency pattern is proposed.Two-level semantic drift suppressing strategy is adopted.Experimental result verifies the AFDDP is a good representation of relations. Relation extraction is essential for most text mining tasks. Existing approaches on relation extraction are generally based on bootstrapping methodology which implies semantic drift problem. This paper presents a new approach to learn semantic dependency patterns, which can significantly alleviate this problem. To this end, a unique representation of activation force defined dependency pattern is presented. It is a trigger word mediated relation between an entity and its attribute value, and the trigger word is extracted by using the statistics of word activation forces between those words. The adaptability and the scalability of the framework are facilitated by the recursive and compositional bootstrap learning of patterns and seed pairs. To obtain insights of the reliability and applicability of the method, we applied it to the English Slot Filling task of Knowledge Base Population track at Text Analysis Conference 2013. Experimental results show that the proposed method has good performance in the implementation of English Slot Filling 2013 with the overall F1 value significantly higher than the best automatic result reported. The experimental results also demonstrate that the activation force based trigger word mining method plays an essential role in improving the performance.

A general formalization of existing bootstrapping frameworks is concluded.A formalization of new semantic bootstrapping model is defined.A unique SSDP to guide learning iterations of bootstrapping.A novel bottom-up kernel method for comparing patterns.The application of the new model to KBP-ESF task. Traditionally, pattern-based relation extraction methods are usually based on iterative bootstrapping model which generally implies semantic drift or low recall problem. In this paper, we present a novel semantic bootstrapping framework that uses semantic information of patterns and flexible match method to address such problem. We introduce formalization for this class of bootstrapping models, which allows semantic constraint to guide learning iterations and use flexible bottom-up kernel to compare patterns. To obtain the insights of reliability and applicability of our framework, we applied it to the English Slot Filling (ESF) task of Knowledge Based Population (KBP) at Text Analysis Conference (TAC). Experimental results show that our framework obtains performance superior to the state of the art.

Nature Language Processing has been paid more attention recently. Traditional approaches for language model primarily rely on elaborately designed features and complicated natural language processing tools, which take a large amount of human effort and are prone to error propagation and data sparse problem. Deep neural network method has been shown to be able to learn implicit semantics of text without extra knowledge. To better learn deep underlying semantics of sentences, most deepneuralnetworklanguagemodelsutilizemulti-gramstrategy. However, the current multi-gram strategies in CNN framework are mostly realized by concatenating trained multi-gram vectors to form the sentence vector, which can increase the number of parameters to be learned and is prone to over fitting. To alleviate the problem mentioned above, we propose a novel adaptive sentence representation learning model based on multigram CNN framework. It learns adaptive importance weights of different n-gram features and forms sentence representation by using weighted sum operation on extracted n-gram features, which can largely reduce parameters to be learned and alleviate the threat of over fitting. Experimental results show that the proposed method can improve performances when be used in sentiment and relation classification tasks.

This paper studies the problem of rural express logistics distribution, considering the dual demand of rural customers, the diversity of agricultural products, the dispersion of rural customer gathering places and so on. Optimize the joint delivery of rural express by introducing the model of vehicle routing problem with split deliveries and pickups (SVRPPD). The objective of optimization is to minimize the total cost of distribution system. According to the characteristics of the model, a competitive decision algorithm is designed to solve the problem. Finally, the practicability and effectiveness of the model and algorithm are verified by an example.

This paper studies the problem of urban express joint distribution. Considering the problem of distribution complexity of city express delivery due to multiple frequency and small quantity and difficult to form economies of scale, we proposed the joint distribution mode and a mathematical model of split delivery vehicle routing problem based on the mode with the goal of minimum total distribution cost. Then an improved competition decision algorithm based on the characteristics of the model was designed. Finally, the effectiveness and practicability of the model and algorithm were illustrated by calculations.

Recently, attention mechanism has been transferred to relation classification task. Since relation classification is a sequence-to-label task, the challenge is how to generate the deciding factor to calculate attention weights. The previous solution randomly initializes a global deciding factor, which is easy to suffer from over-fitting. To solve the problem, we propose instance-adaptive attention mechanism, which generates a specially designed deciding factor for each sentence. The experimental result on SemEval-2010 Task 8 dataset shows that our method can outperform most state-of-the-art systems without external linguistic features.

Harmful algal blooms recently have been under the spotlight throughout the world, because of their negative impact on the marine environment, aquaculture, fisheries as well as public health. The development of methods for rapid and precise identification and quantification of causative species is essential for the warning andmonitoring of blooms, among which the techniques based on taxonomic probes are themost favored. In this study, two harmful algae, i.e., Prorocentrum minimum and Karenia mikimotoi were taken into consideration. The partial large subunit rDNA (D1-D2) of both species were firstly PCR-amplified, cloned and sequenced. The obtained sequences were then introduced to carry out alignment analysis for gene specific regions. Three respective candidate probes for each species were designed and used to screen the optimal probe by performing fluorescence in situ hybridization (FISH) tests. The results showed that the probes Pmin0443 and Kmik0602 displayed the best hybridization for P. minimum and K. mikimotoi, respectively. Both the specific (taxonomic) (Pmin0443 and Kmik0602) and the control probes (UniC0512 and UniR0499) were used for cross-reactivity tests with other microalgae in our laboratory. The probes Pmin0443 and Kmik0602 are specific and could be served as taxonomic probes introduced into the techniques targeting rRNA, such as FISH, sandwich hybridization, and DNA-microarray assay of P. minimum and K. mikimotoi in the future. Finally, FISH analyses with both probes were performed on the simulated field samples. The probes could hybridize exclusively with the target cells well, and no significant difference (p > 0.05) was observed in the cell densities of the samples determined by FISH and light microscopy (LM). All suggest that the probes are specific and could be introduced into FISH for the monitoring of both harmful algae.

The outbreak of unexpected news events such as large human accident or natural disaster brings about a new information access problem where traditional approaches fail. Mostly, news of these events shows characteristics that are early sparse and later redundant. Hence, it is very important to get updates and provide individuals with timely and important information of these incidents during their development, especially when being applied in wireless and mobile Internet of Things (IoT). In this paper, we define the problem of sequential update summarization extraction and present a new hierarchical update mining system which can broadcast with useful, new, and timely sentence-length updates about a developing event. The new system proposes a novel method, which incorporates techniques from topic-level and sentence-level summarization. To evaluate the performance of the proposed system, we apply it to the task of sequential update summarization of temporal summarization (TS) track at Text Retrieval Conference (TREC) 2013 to compute four measurements of the update mining system: the expected gain, expected latency gain, comprehensiveness, and latency comprehensiveness. Experimental results show that our proposed method has good performance.

Measuring the similarity of patterns is the key in pattern-based approaches in relation extraction. Most existing methods generally rely on inflexible pattern similarity measurements which often lead to low recall. In this work, a novel kernel-based model is proposed to address this problem. Depending on the pattern similarities produced by our bottom-up kernel, the most similar semantic shortest dependency patterns are selected to update seed patterns in each iteration of bootstrapping. To obtain insights of the reliability and applicability of our proposed method, we applied it to the task of English Slot Filling (ESF) in Knowledge Base Population (KBP) track at Text Analysis Conference (TAC). The experimental results validate our proposed method that importantly improves the recall which resulting in the improvement of F1 value. The effectiveness of the bottomup kernel is also verified by further experimental results.

Ventricular fibrillation (VF) is a malignant arrhythmia, which belongs to complex and nonlinear signals. To realize the detection of ventricular fibrillation, a new algorithm, which was based on the improved threshold crossing interval (TCI) algorithm and Support vector machine (SVM), was proposed in this paper. The SWM has great advantages in processing classification and pattern recognition. In this paper, 4-s-sliding-window technology and the improved TCI algorithm are applied to extract features of electrocardiogram (ECG). The improved TCI algorithm was implemented as follows: firstly, the average threshold crossing interval value of the middle 2s was calculated by using absolute threshold at each 4-s-sliding-window; secondly, the feature parameter (TCI value) was input to a pre-designed binary classification support vector machine; lastly, the classification was accomplished. For evaluating the reliability of the new algorithm, both MIT-BIH arrhythmia database and CU ventricular tachyarrhythmia database were used to test. By comparing the sensitivity, specificity, positive predictability and accuracy with other well known methods, the conclusion was made that this method is superior to other methods. This new algorithm is easier to implement and has greater advantages in real-time execution. These advantages make it more suitable in real time ECG monitoring and defibrillator.

An algorithm for detecting ventricular fibrillation (VF) by the method of support vector machine is presented. The algorithm first extracts the feature of electrocardiogram in every 4 s sliding window by the improved time delay method and the parameter d is obtained as feature; the support vector machine method is used to realize the discrimination of VF and non-VF signals. For evaluating the new algorithm, the complete BIH-MIT arrhythmia database and the CU database were used to simulate without any pre-selection. The sensitivity, specificity, positive predictability and accuracy were calculated and compared these values with results from an earlier investigation of several different ventricular fibrillation detection algorithms. It shows that the new algorithm has good performance and has greater advantages in real-time execution.