Drug-target interaction (DTI) prediction has an important role in drug discovery, significantly expediting the drug design and development process. This paper proposes a new model, ERW_BiAN, that seamlessly integrates graph embedding representations with deep learning network. Specifically, we utilize an improved graph embedding model, ERW, to generate comprehensive feature vectors for each node within the knowledge graph. Subsequently, these feature vectors are fed into the BiAN model, which leverages an attention mechanism to assign precise weights to sequences. Experimental results demonstrate the superior accuracy and predictive prowess of ERW_BiAN in DTI prediction tasks. Furthermore, the case about COVID-19 shows that ERW_BiAN has the better application potential for predicting drug-target interactions.

Summary High penetration of renewable energy is the development trend of the future power system. As one of the clean energy sources, wind power generation has an increasing share in the energy market. However, due to the harsh working environment, the high fault rate and poor accessibility of the wind farms, resulting in the difficult maintenance process and high cost. This article proposes a fault diagnosis (FD) method based on long short‐term memory (LSTM) and feature optimization strategies for wind turbines (WTs), thus reducing the operation and maintenance costs of WTs. First, Pearson correlation coefficient analysis is performed on the collected data features to remove redundant features, and wavelet transform is adopted to remove the redundant data, so as to optimize the fault features and fault data. Then the selected features samples are used to train LSTM‐based FD model. Finally, the actual production data is adopted to verify the proposed method. The proposed method can effectively locate the faults, and provide data support for wind farms, thus improving the reliability, safety, and economic benefits of wind farms.

Although much progress has been made in new drug development, it is still a costly, complicated and less efficient process. Therefore, drug repositioning studies are highly desirable. So far, many methods based on known drug-target interactions (DTIs) have been designed to detect potential DTIs, but there are many challenges for improving the performance of prediction DTIs. This paper proposes a new method (CaGCN-DTI) for DTIs prediction, which uses a heterogeneous network that incorporates a wide variety of biological data (drug, target, disease, and side effect) to discover potential interactions between drugs and targets. First, the drug-protein similarity network is preprocessed by Spectral clustering, then the graph convolutional network (GCN) combined with attention mechanism and random walk with restart (RWR) is used to aggregate message transmission to the network. Finally, the embedding vectors are used to discover potential DTIs by matrix decomposition. Experiments are performed based on ten-fold cross-validation, and the results show that CaGCN-DTI outperforms other previous methods in terms of AUC and AUPR.

The prediction of drug-target interactions (DTIs) is of great significance to the fields of drug design and drug development. However, traditional biological experiments are time-consuming and cost-effective, which has prompted more people to turn their attention to the use of computers to assist in predicting DTIs. This paper proposes an improved prediction model based on multiple graph representation methods, which is GDNet-DTI that combined GCN and DeepWalk. First, a molecular map with atoms as nodes and chemical bonds as edges is generated using the SMILE sequence of drugs, and then GIN is used to extract the features of molecular map for better obtaining the complex interactions between atoms. For target proteins, the protein sequence is first represented by a word vector, and then the one-dimensional convolution is used to extract features for extracting the different levels of features. Then, based on obtained drug features and target features, a DTI-graph is generated, in which drugs and targets are represented as nodes and interactions are represented as edges. Finally, GDNet-DTI are used to obtain node neighborhood information and graph topology information of the DTI-graph. Compared with other advanced models, the results show that GDNet-DTI combined with multiple graph features can predict DTIs more accurately and effectively with DrugBank and four benchmark datasets. In addition, a case study with COVID-19 data is presented, which shows that the proposed method has the potential to predict the actual DTIs and can contribute to the development of drug discovery.

In practical application, the existing gender and age recognition algorithms can’t meet the requirements of both smallsized model and high accuracy simultaneously. Moreover, most models based on CNNs have a even larger size of more than 200M. In this paper a multi-layer multi-channel attentive network based on the idea of divide-and-conquer is proposed. This method uses multi-layer processes in each channel to extract features of different layers and improves accuracy by layer refinement. We use some dynamic parameters to fine- tune each layer to make the model fit better. Each layer uses the same classifier to reduce parameters to make the model smaller. And we import attention mechanisms to increase the ability of the network to use the features. Experiments show that the accuracy of this method is better than several mainstream networks and the size of the model is less than 0.5M, which can be used in mobile terminals well.

Trust Relationships have shown great potential to improve recommendation quality, especially for cold start and sparse users. Since each user trust their friends in different degrees, there are numbers of works been proposed to take Trust Strength into account for recommender systems. However, these methods ignore the information of trust directions between users. In this paper, we propose a novel method to adaptively learn directive trust strength to improve trust-aware recommender systems. Advancing previous works, we propose to establish direction of trust strength by modeling the implicit relationships between users with roles of trusters and trustees. Specially, under new trust strength with directions, how to compute the directive trust strength is becoming a new challenge. Therefore, we present a novel method to adaptively learn directive trust strengths in a unified framework by enforcing the trust strength into range of [0, 1] through a mapping function. Our experiments on Epinions and Ciao datasets demonstrate that the proposed algorithm can effectively outperform several state-of-art algorithms on both MAE and RMSE metrics.

To study the failure mechanism of LED driver and predict the life of it, this paper analyzes the LED driver possible failure forms and the causes which make the driver fail. Three accelerated life model are introduced in this paper and how to get the parameter is discussed. LED diver high accelerated life testing platform is built to test the sample driver life and get the parameter in the accelerated life model. Three different application of constant temperature and humidity mixed environment test are used to stress the sample LED driver. The accelerated life model estimate parameter is obtained and the classical forms of LED driver is sudden appearance style. Finally the life of sample LED driver is estimated by the model introduced above.