Efficient Multi-modal Hashing with Online Query Adaption for Multimedia Retrieval

Multi-modal hashing supports efficient multimedia retrieval well. However, existing methods still suffer from two problems: (1) Fixed multi-modal fusion. They collaborate the multi-modal features with fixed weights for hash learning, which cannot adaptively capture the variations of online streaming multimedia contents. (2) Binary optimization challenge. To generate binary hash codes, existing methods adopt either two-step relaxed optimization that causes significant quantization errors or direct discrete optimization that consumes considerable computation and storage cost. To address these problems, we first propose a Supervised Multi-modal Hashing with Online Query-adaption method. A self-weighted fusion strategy is designed to adaptively preserve the multi-modal features into hash codes by exploiting their complementarity. Besides, the hash codes are efficiently learned with the supervision of pair-wise semantic labels to enhance their discriminative capability while avoiding the challenging symmetric similarity matrix factorization. Further, we propose an efficient Unsupervised Multi-modal Hashing with Online Query-adaption method with an adaptive multi-modal quantization strategy. The hash codes are directly learned without the reliance on the specific objective formulations. Finally, in both methods, we design a parameter-free online hashing module to adaptively capture query variations at the online retrieval stage. Experiments validate the superiority of our proposed methods.

A Credit Conflict Detection Model Based on Decision Distance and Probability Matrix

Considering the credit index calculation differences, semantic differences, false data, and other problems between platforms such as Internet finance, e-commerce, and health and elderly care, which lead to the credit deviation from the trusted range of credit subjects and the lack of related information of credit subjects, in this paper, we proposed a crossplatform service credit conflict detection model based on the decision distance to support the migration and application of crossplatform credit information transmission and integration. Firstly, we give a scoring table of influencing factors. Score is the probability of the impact of this factor on credit. Through this probability, the distance matrix between influencing factors is generated. Secondly, the similarity matrix is calculated from the distance matrix. Thirdly, the support vector is calculated through the similarity matrix. Fourth, the credit vector is calculated by the support vector. Finally, the credibility is calculated by the credit vector and probability.

Fusion of Internal Similarity to Improve the Accuracy of Recommendation Algorithm

Collaborative filtering algorithms (CF) and mass diffusion (MD) algorithms have been successfully applied to recommender systems for years and can solve the problem of information overload. However, both algorithms suffer from data sparsity, and both tend to recommend popular products, which have poor diversity and are not suitable for real life. In this paper, we propose a user internal similarity-based recommendation algorithm (UISRC). UISRC first calculates the item-item similarity matrix and calculates the average similarity between items purchased by each user as the user’s internal similarity. The internal similarity of users is combined to modify the recommendation score to make score predictions and suggestions. Simulation experiments on RYM and Last.FM datasets, the results show that UISRC can obtain better recommendation accuracy and a variety of recommendations than traditional CF and MD algorithms.

LE-MDCAP: A Computational Model to Prioritize Causal miRNA–Disease Associations

MicroRNAs (miRNAs) are associated with various complex human diseases and some miRNAs can be directly involved in the mechanisms of disease. Identifying disease-causative miRNAs can provide novel insight in disease pathogenesis from a miRNA perspective and facilitate disease treatment. To date, various computational models have been developed to predict general miRNA–disease associations, but few models are available to further prioritize causal miRNA–disease associations from non-causal associations. Therefore, in this study, we constructed a Levenshtein-Distance-Enhanced miRNA–Disease Causal Association Predictor (LE-MDCAP), to predict potential causal miRNA–disease associations. Specifically, Levenshtein distance matrixes covering the sequence, expression and functional miRNA similarities were introduced to enhance the previous Gaussian interaction profile kernel-based similarity matrix. LE-MDCAP integrated miRNA similarity matrices, disease semantic similarity matrix and known causal miRNA–disease associations to make predictions. For regular causal vs. non-disease association discrimination task, LF-MDCAP achieved area under the receiver operating characteristic curve (AUROC) of 0.911 and 0.906 in 10-fold cross-validation and independent test, respectively. More importantly, LE-MDCAP prominently outperformed the previous MDCAP model in distinguishing causal versus non-causal miRNA–disease associations (AUROC 0.820 vs. 0.695). Case studies performed on diabetic retinopathy and hsa-mir-361 also validated the accuracy of our model. In summary, LE-MDCAP could be useful for screening causal miRNA–disease associations from general miRNA–disease associations.

Morphological characterization of natural orchids Dendrobium spp.

Dendrobium is characterized by long pseudobulbs or canes with soft leaves over the entire length, or in some species short or swollen pseudobulbs with two leathery leaves. The inflorescence is composed from dozens of flowers of different sizes and colors. This study aimed to identify the quantitative morphological character of five species of Dendrobium spp. namely D. mirbelianum, D. lamellatum from Java, D. anosmum from South Kalimantan, D. bracteosum from Papua, and D. purpureum from North Sumatera. The resulted dendrogram based on the similarity matrix were divided into two clusters, among the five species the value of similarity coefficient is 1.50. The first cluster is only composed from D. mirbelianum, the second cluster is D. lamellatum, D. purpureum, D. bracteosum, and D. anosmum which have more distant relationship with the other three orchids. Moreover, D. lamellatum and D. purpureum have the closest similarity coefficient with 0.81 value, which have bigger chance to use as the parents for hybridization. There are many Dendrobiums spp. distribution which based on the relationship area. In addition to quantitative properties, it also needs to be expanded to qualitative, anatomy, cytology, and also molecular characteristic to have more comprehensive data.

A Method for Recommending Sports Network Course Resources Based on Collaborative Filtering Technology

In the process of online course resource recommendation, the output of recommendation results is often unstable. Therefore, a physical education online course resource recommendation method based on collaborative filtering technology is proposed. Firstly, the learning preference of e-learners is calculated, the frequency index of the word frequency-inverse document is defined, the correlation between courses is reflected, and the specific needs of students for PE online course resource recommendation are understood. Then, the collaborative filtering recommendation algorithm is used to generate the similarity matrix and correlation matrix, update the edge characteristics of sports online curriculum resources, collect and refine the hidden index of sports online curriculum resources, optimize the prediction rules of the neighborhood of the most similar teaching unit, and complete the recommendation of sports online curriculum resources. Experimental results show that, for 1000 keywords, the method has the highest single average matching degree, the recommendation process is stable, and the F1 value is more than 0.9, and the practical application is ensured.

MULTIMODAL HUMAN BRAIN LONGITUDINAL PARCELLATION ACROSS LIFESPAN

Subdividing the human brain into several functionally distinct and spatially contiguous areas is important to understand the amazingly complex human cerebral cortex. However, adult aging is related to differences in the structure, function, and connectivity of brain areas, so that the single population subdivision does not apply to multiple age groups. Moreover, different modalities could provide affirmative and complementary information for the human brain subdivision. To obtain a more reasonable subdivision of the cerebral cortex, we make use of multimodal information to subdivide the human cerebral cortex across lifespan. Specifically, we first construct a population average functional connectivity matrix for each modality of each age group. Second, we separately calculate the population average similarity matrix for the cortical thickness and myelin modality of each age group. Finally, we fuse these population average matrixes to obtain the multimodal similarity matrix and feed it into the spectral clustering algorithm to generate the brain parcellation for each age group.

Discrete Two-Step Cross-Modal Hashing through the Exploitation of Pairwise Relations

The cross-modal hashing method can map heterogeneous multimodal data into a compact binary code that preserves semantic similarity, which can significantly enhance the convenience of cross-modal retrieval. However, the currently available supervised cross-modal hashing methods generally only factorize the label matrix and do not fully exploit the supervised information. Furthermore, these methods often only use one-directional mapping, which results in an unstable hash learning process. To address these problems, we propose a new supervised cross-modal hash learning method called Discrete Two-step Cross-modal Hashing (DTCH) through the exploitation of pairwise relations. Specifically, this method fully exploits the pairwise similarity relations contained in the supervision information: for the label matrix, the hash learning process is stabilized by combining matrix factorization and label regression; for the pairwise similarity matrix, a semirelaxed and semidiscrete strategy is adopted to potentially reduce the cumulative quantization errors while improving the retrieval efficiency and accuracy. The approach further combines an exploration of fine-grained features in the objective function with a novel out-of-sample extension strategy to enable the implicit preservation of consistency between the different modal distributions of samples and the pairwise similarity relations. The superiority of our method was verified through extensive experiments using two widely used datasets.