Extracting entities and relations is crucial for building large-scale knowledge graph and different knowledge extraction tasks. Based on the pre-trained language model, an entity-oriented joint extraction method combining subject attention was proposed. In this method, the key information of the subject was extracted by using Convolutional Neural Network (CNN) and the dependency relationship between the subject and the object was captured by the attention mechanism. Followed by the above, a Joint extraction model based on Subject Attention (JSA) was built. In experiments on public dataset New York Times corpus (NYT) and the dataset of artificial intelligence built by distant supervision, the F1 score of the proposed model was improved by 1.8 and 8.9 percentage points respectively compared with Cascade binary tagging framework for Relational triple extraction (CasRel).

On-line detection of fabric defects is a major problem faced by textile industry. Aiming at the problems such as high false positive rate, high false negative rate and low real-time in the existing detection of fabric defects, an on-line detection algorithm for fabric defects based on deep learning was proposed. Firstly, based on GoogLeNet network architecture, and referring to classical algorithm of other classification models, a fabric defect classification model suitable for actual production environment was constructed. Secondly, a fabric defect database was set up by using different kinds of fabric pictures marked by quality inspectors, and the database was used to train the fabric defect classification model. Finally, the images collected by high-definition camera on fabric inspection machine were segmented, and the segmented small images were sent to the trained classification model in batches to realize the classification of each small image. Thereby the defects were detected and their positions were determined. The model was validated on a fabric defect database. The experimental results show that the average test time of each small picture is 0.37 ms by this proposed model, which is 67% lower than that by GoogLeNet, 93% lower than that by ResNet-50, and the accuracy of the proposed model is 99.99% on test set, which shows that its accuracy and real-time performance meet actual industrial demands.

Since the present performance of smartphone based personal indoor localization is still far from satisfaction with problems in accuracy, cost and etc., a duel-step filtering indoor localization algorithm fusing Wi-Fi fingerprints and images implemented on devices like smartphones was proposed in this paper. This algorithm comprised of offline stage and online stage. On the offline stage, Wi-Fi fingerprints were collected and a fingerprint library sampled in different positions in the coordinate system was constructed. Photos were taken in that stage to extract the image features. On the online stage, the first filter process was to determine the possible area where the user is currently by using Wi-Fi information captured in real-time. Then a distance compensation algorithm was proposed in this paper to extract features of the real-time image taken by the user to determine the exact localized position. Experimental results show that this algorithm can effectively improve localization precision compared with traditional Wi-Fi and image based localization methods in environments with fewer APs (Access Points) and similar layouts, thus is capable for general localization or LBS (Location-Based Service) relevant applications.

In the distributed storage system, Disk space Utilization (DU) is generally used to measure the load balance of each storage node. When given the equal disk space utilization to each node, the balance of storage load is achieved in the whole distributed storage system. However, in practice, the storage node with relatively low disk I/O speed and reliability becomes a bottleneck for the performance of data I/O in the whole storage system. Therefore in heterogeneous distributed storage system and specially the system which has great differences in disk I/O speed and reliability of each storage node, the speed of data I/O is definitely limited when disk space utilization is the only evaluation criteria of storage load balance. A new idea based on read-write efficiency was proposed to measure the storage load balance in the distributed storage system. According to the definition of Storage Entropy (SE) given by the theory of load balance and entropy, a kind of load balance algorithm based on SE was proposed. With system load and single node load determination as well as load shifting, the quantitative adjustment for storage load of the distributed storage system was achieved. The proposed algorithm was tested and compared with the load balance algorithm based on disk space utilization. Experimental results show that the proposed algorithm can balance storage load well in the distributed storage system, which effectively restrains the system load imbalance and improves the overall efficiency of reading and writing of the distributed storage system.

The construction of structure is the foundation of indoor map constructing, and the indoor distance measuring is one of the core problems in this process. In order to solve the problem of high cost or low accuracy in the existing methods, a distance measuring method based on the statistical steps and strides with the multi-sensor data was proposed. In the stage of counting steps, the optimal threshold was calculated according to the ideas of Support Vector Machine (SVM), which made the model have excellent generalization ability. In the stage of detecting the validity of the step, the variance of the direction sensor data was used to filter the effective displacement steps. Finally, the stride estimation model was used to estimate the stride, and then the effective displacement was calculated. In the practical application, the proposed method is proved to be effective, and the overall error is about 4%, which can achieve the accuracy required to build indoor maps. It is a low cost and reliable displacement calculation method for indoor map constructing.

In ranking for document retrieval, queries often vary greatly from one another. However, most of the existing ranking methods do not consider significant differences between queries. Correctly ranking documents on the top of the result list is crucial, and one must conduct training in a way that such ranked results are accurate. A two-tier weighting aggregation ranking method was proposed. This method consisted of two steps, training of base rankers and query-level ranker aggregation. First, base rankers were established based on each query, assigning asymmetric weights to its relevant documents, then, query-level ranker aggregation used a supervised approach to learn query-dependent weights when these base rankers were combined. The experimental results on the benchmark data set LETRO ONHSUMED show that the ranking performance has been significantly improved.

Concerning the parameter optimization of electric assist control strategy for Hybrid Electrical Vehicles (HEV), the orthogonal table was designed based on the theory of orthoplan experiment, which took the fuel consumption, the COs emission, and the total emission of HC and NOx as the experimental indexes. The influence of control strategy parameters on the parallel HEV was obtained by using intuitive analysis to analyze the 18 group test results, and then the remarkably influencing factors of each index were found out.

In order to increase the probability of population-based search approach to capture specific objects from the decision space within limited time, based on the classical probability model, this paper established an overall random sampling model and a partition sampling model inspired by stratified sampling which divided the sample space into more than one subspace. By analyzing and comparing the two random event's probability of obtaining the specific objectives at least one time among repeatedly independent random sampling from those models, the paper proves that the partition sampling model's probability is greater than the overall random sampling model's probability permanently when the amount of specific objective in the collectivity is 1 or 2.

Weight coefficients should be employed to transform multi-objective problem of hybrid system into a single objective one. In order to avoid setting weight coefficients, a methodological approach based on Strength Pareto Evolutionary Algorithm (SPEA2) was proposed to optimize parameters of constrained Parallel Hybrid Electric Vehicle (PHEV).The Pareto dominance principle was employed to judge candidate solutions and the objective was to minimum fuel consumption and exhaust emissions while ADVISOR was used to simulate the PHEV driving. The optimal results demonstrate that adopting the methodological approach proposed in this paper to optimize parameters of power control strategy and drivetrain has a significant effect on enhancing working efficiency, promoting vehicle performance, decreasing fuel consumption and reducing exhaust emissions of PHEV.