Understanding the meaning of mathematical problems is the key for automatic problem solving. However， the accuracy of understanding word problems with complex situations and many parameters is relatively low in previous studies， and the effective optimization solutions need to be further explored and studied. On this basis， a math word problem understanding method integrating commonsense knowledge base and grammatical features was proposed for the classical probability word problems with complex context. Firstly， a classical probability word problem representation model containing seven kinds of key problem-solving parameters was constructed according to text and structure characteristics of the classical probability word problems. Then， based on this model， the task of understanding of word problems was transformed into the problem of solving parameter identification， and a Conditional Random Field （CRF） parameter identification method integrating multi-dimensional grammatical features was presented to solve it. Furthermore， aiming at the problem of implicit parameter identification， a commonsense completion module was added， and an understanding method of math word problems integrating commonsense knowledge base and grammatical features was proposed. Experimental results show that the proposed method has the average F1-score of 93.56% for problem-solving parameter identification， and the accuracy of word problem understanding reached 66.54%， which are better than those of Maximum Entropy Model （MaxEnt）， Bidirectional Long Short-Term Memory-Conditional Random Field （BiLSTM-CRF） and traditional CRF methods. It proves the effectiveness of this method in understanding of classical probability word problems.

Aiming at the data noise filtering problem and the importance problem of different POIs in POI (Point-Of-Interest)recommendation research, a POI recommendation algorithm, named RecSI (Recommendation by Spatiotemporal information and POI Importance), was proposed. First, the geographic information and the mutual attraction between the POIs were used to filter out the data noise, so as to narrow the range of candidate set. Second, the user’s preference score was calculated by combining the user’s preference on the POI category at different time periods of the day and the popularities of the POIs. Then, the importances of different POIs were calculated by combining social information and weighted PageRank algorithm. Finally, the user’s preference score and POI importances were linearly combined in order to recommend TOP- K POIs to the user. Experimental results on real Foursquare sign-in dataset show that the precision and recall of the RecSI algorithm are higher than those of baseline GCSR (Geography-Category-Socialsentiment fusion Recommendation) algorithm by 12.5% and 6% respectively, which verify the effectiveness of RecSI algorithm.

Concerning the problem that the accuracy and real-time effects of virtual-real registration in Augmented Reality (AR) based on vision are greatly affected by the changes of illumination, occlusion and perspective, which is easy to lead to failure of registration, a virtual-real registration method of natural features based on Binary Robust Invariant Scalable Keypoints-Speeded Up Robust Features (BRISK-SURF) algorithm was proposed. Firstly, Speeded Up Robust Features (SURF) feature extractor was used to detect the feature points. Then, Binary Robust Invariant Scalable Keypoints (BRISK) descriptor was used to describe the feature points in binary, and the feature points were matched accurately and efficiently by combining Hamming distance. Finally, the virtual-real registration was realized according to the homography relationship between images. Experiments were performed from the aspects of image feature matching and virtual-real registration. Results show that the average precision of BRISK-SURF algorithm is basically the same with that of SURF algorithm, is about 25% higher than that of BRISK algorithm, and the average recall of BRISK-SURF is increased by about 10% compared to that of BRISK algorithm; the result of the virtual-real registration method based on BRISK-SURF is close to the reference standard data with high precision and good real-time performance. The Experimental results illustrate that the proposed method has high recognition accuracy, registration precision and real-time effects for images with different illuminations, occlusions and perspectives. Besides, the interactive tourist resource presentation and experience system based on AR is realized by using the proposed method.

The large number of duplicate images in the database not only affects the performance of the learner, but also consumes a lot of storage space. For massive image deduplication, a duplicate detection algorithm for massive images was proposed based on pHash (perception Hashing). Firstly, the pHash values of all images were generated. Secondly, the pHash values were divided into several parts with the same length. If the values of one of the pHash parts of the two images were equal to each other, the two images might be duplicate. Finally, the transitivity of image duplicate was discussed, and corresponding algorithms for transitivity case and non-transitivity case were proposed. Experimental results show that the proposed algorithms are effective in processing massive images. When the similarity threshold is 13, detecting the duplicate of nearly 300000 images by the proposed transitive algorithm only takes about two minutes with the accuracy around 53%.

In view of the optimization of high-speed train connection in passenger transport hub under the condition of high-speed railway network, the concept of transfer satisfaction of medium and long distance passenger flow was proposed by analyzing the passenger transfer process in hub, and a high-speed train connection optimization model for large passenger transport hub based on transfer orientation was proposed with the average transfer satisfaction and the arrival and departure equilibrium of trains at hub stations as the optimization objective and with the constraint conditions of reasonable originating time of large stations, reasonable terminating time, station operation interval time, passenger transfer time and station arrival and departure line capacity. A genetic algorithm with improved chromosome coding mode and selection strategy was designed to solve the example. Compared with the basic genetic algorithm and the basic simulated annealing algorithm, the improved genetic algorithm increases the average transfer satisfaction in the objective function by 5.10% and 2.93% respectively, and raises the equilibrium of arrival and departure of trains at hub stations by 0.27% and 2.31% respectively. The results of the example verify the effectiveness and stability of the improved genetic algorithm, which indicates that the proposed method can effectively optimize the quality of the high-speed train connection in large passenger transport hub.

In order to improve the quality and efficiency of railway crew routing plan, the problem of crew routing plan was abstracted as a Multi-Traveling Salesman Problem (MTSP) with single base and balanced travel distance, and a equilibrium factor was introduced to establish a mathematical model aiming at less crew routing time and balanced tasks between sub-crew routings. A dual-strategy ant colony optimization algorithm was proposed for this model. Firstly, a solution space satisfying the space-time constraints was constructed and pheromone concentration was set for the node of the crew section and the continuation path respectively, then the transitional probability of the dual-strategy state was adopted to make the ant traverse all of the crew segments, and finally the sub-crew routings that meet the crew constraint rules were found. The designed model and algorithm were tested by the data of the intercity railway from Guangzhou to Shenzhen. The comparison with the experimental results of genetic algorithm shows that under the same model conditions, the number of crew routing in the crew routing plan generated by double-strategy ant colony optimization algorithm is reduced by about 21.74%, the total length of crew routing is decreased by about 5.76%, and the routing overload rate is 0. Using the designed model and algorithm to generate the crew routing plan can reduce the crew routing time of crew plan, balance the workload and avoid overload routing.

Focusing on the issue that emergency resource assignment for multiple demand points and multiple supply points in railway emergencies, an emergency resource assignment model of multiple rescue targets was established, which was based on the concept of "soft time window". The maximum fairness and minimum total assignment cost were considered as the optimization objectives, and parallel selected genetic algorithm was used to solve the model. The population was equally divided into subpopulations by the algorithm. Subpopulations' number was equal to the number of objective functions. An objective function was assigned to each divided subpopulation and the selection work was done independently, by which individuals with high fitness were selected from each subpopulation to form a new population. Crossover and mutation were done to generate the next generation of population. The computing cases show that the parallel selected genetic algorithm reduces the variance of resource satisfaction degree of all demand points by 93.88% and 89.88% respectively, and cuts down the cost by 5% and 0.15% respectively, compared with Particle Swarm Optimization (PSO) and two-phase heuristic algorithm. The proposed algorithm can effectively reduce the variance of the resource satisfaction degree of all demand points, that is, it improves the fairness of each demand point and reduces the cost at the same time, and can obtain higher quality solution when solving multiple objective programming problem.

Aiming at the problem that the traditional Laplacian score for feature selection cannot be directly applied to multi-label tasks, a multi-label feature selection algorithm based on Laplacian score was proposed. Firstly, the sample similarity matrix was reconstructed by the correlation of the common and non-correlated correlations of the samples in the overall label space. Then, the correlation and redundancy between features were introduced into Laplacian score, and a forward greedy search strategy was designed to evaluate the co-operation ability between candidate features and selected features, which was used to evaluate the importance of candidate features. Finally, extensive experiments were conducted on six multi-label data sets with five different evaluation criteria. The experimental results show that compared with Multi-label Dimensionality reduction via Dependence Maximization (MDDM), Feature selection for Multi-Label Naive Bayes classification (MLNB) and feature selection for multi-label classification using multivariate mutual information (PMU), the proposed algorithm not only has the best classification performance, but also has a remarkable performance of up to 65%.

Aiming at the problems of scalability, algorithm convergence performance and operational efficiency in building large-scale machine learning systems, the challenges of the large-scale sample, model and network communication to the machine learning system were analyzed and the solutions of the existing systems were also presented. Taking Latent Dirichlet Allocation (LDA) model as an example, by comparing three open source distributed LDA systems-Spark LDA, PLDA+ and LightLDA, the differences in system design, implementation and performance were analyzed in terms of system resource consumption, algorithm convergence performance and scalability. The experimental results show that the memory usage of LightLDA and PLDA+ is about half of Spark LDA, and the convergence speed is 4 to 5 times of Spark LDA in the face of small sample sets and models. In the case of large-scale sample sets and models, the network communication volume and system convergence time of LightLDA is much smaller than PLDA+ and SparkLDA, showing a good scalability. The model of "data parallelism+model parallelism" can effectively meet the challenge of large-scale sample and model. The mechanism of Stale Synchronous Parallel (SSP) model for parameters, local caching mechanism of model and sparse storage of parameter can reduce the network cost effectively and improve the system operation efficiency.

For a better trade-off between fault-tolerance mechanism and fault-tolerance overhead in processor reliability research, a fault detection approach for Multi-Processor System-on-Chip (MPSoC) that placed the calculation task of detecting code on redundancy core was proposed in this paper. The approach achieved MPSoC failure detection by placing the calculation and comparison parts of detecting code on redundancy core. The technique required no additional hardware modification, and shortened the design cycle while reducing performance and memory overheads. The verification experiment was implemented on a MPSoC by fault injection and running multiple benchmark programs. Comparing several previous methods of fault detection in terms of capability, area, memory and performance overhead, the experiment results show that the approach is effective and able to achieve a better trade-off between performance and overhead.