Knowledge Graph (KG) strongly support the research of knowledge-driven artificial intelligence. Aiming at this fact, the existing technologies of knowledge graph and knowledge hypergraph were analyzed and summarized. At first, from the definition and development history of knowledge graph, the classification and architecture of knowledge graph were introduced. Second, the existing knowledge representation and storage methods were explained. Then, based on the construction process of knowledge graph, several knowledge graph construction techniques were analyzed. Specifically, aiming at the knowledge reasoning, an important part of knowledge graph, three typical knowledge reasoning approaches were analyzed, which are logic rule-based, embedding representation-based, and neural network-based. Furthermore, the research progress of knowledge hypergraph was introduced along with heterogeneous hypergraph. To effectively present and extract hyper-relational characteristics and realize the modeling of hyper-relation data as well as the fast knowledge reasoning, a three-layer architecture of knowledge hypergraph was proposed. Finally, the typical application scenarios of knowledge graph and knowledge hypergraph were summed up, and the future researches were prospected.

Duing to the complicated pore structure of low porosity and low permeability reservoirs, the prediction accuracy of the existing Nuclear Magnetic Resonance (NMR) logging permeability model for low porosity and low permeability reservoirs is not high. In order to solve the problem, a permeability prediction method based on Deep Belief Network (DBN) algorithm and Kernel Extreme Learning Machine (KELM) algorithm was proposed. The pre-training of DBN model was first carried out, and then the KELM model was placed as a predictor in the trained DBN model. Finally, the Deep Belief Kernel Extreme Learning Machine Network (DBKELMN) model was formed with supervised training by using the training data. Considering that the proposed model should make full use of the information of the transverse relaxation time spectrum which reflected the pore structure, the transverse relaxation time spectrum of NMR logging after discretization was taken as the input, and the permeability was taken as the output. The functional relationship between the transverse relaxation time spectrum of NMR logging and permeability was determined, and the reservoir permeability was predicted based on the functional relationship. The applications of the example show that the permeability prediction method based on DBN algorithm and KELM algorithm is effective and the Mean Absolute Error (MAE) of the prediction sample is 0.34 lower than that of Schlumberger Doll Researchcenter (SDR) model. The experimental results show that the combination of DBN algorithm and KELM algorithm can improve the prediction accuracy of low porosity and low permeability reservoir, and can be used to the exploration and development of oil and gas fields.

Concerning that there is no research about the relationship between disease classification and functional homogeneity analysis of functional protein module in network medicine, the following research work was carried out. Firstly, a gene relationship network was constructed based on the Mesh database and String9 database. Secondly, the gene relationship network was divided by using optimized modularity-based module classification method (such as BGLL, Nonnegtive Matrix Factorization (NMF) and other clustering algorithms). Thirdly, the GO enrichment analysis was carried out for divided modules, and through the comparison of GO enrichment analysis to the high and low pathogenic topology module, important biology suggests for disease classification could be found from protein functional module characteristics in the aspects of biological process, cellular component, molecular function and so on. Finally, the functional characteristics of topological module for disease classification were analyzed, and the data about the functional features of each module was obtained by the analysis to the properties of the network topology such as average degree, density, and average shortest path length, and further correlativity between disease classification and functional module was revealed.

The physical condition of miners escaping from a mine fire in the harmful circumstance is critical to the success of escape. This paper proposed the concept of health degree of escaping miners and analyzed the effect of each harmful factor. A model was built to evaluate all factors' influence on escaping miners based on fuzzy comprehensive evaluation approach, and then a dynamic health degree evaluation method of miners escaping from a mine fire was proposed. Fire Dynamics Simulator (FDS) software was used to simulate a simplified mine fire, and escaping miner's health degree was calculated using the method. The rationality of the method was verified by the experiment. Miner's health degree can evaluate miner's physical state in complex disaster environment synthetically, and it provides a quantifiable basis to guide the decision-making of escape path in the underground disasters.

Most of existing manifold learning algorithms are not capable of dealing with new arrival samples. Although some incremental algorithms are developed via extending a specified manifold learning algorithm, most of them have some disadvantages more or less. In this paper, a novel and more Generalized Incremental Manifold Learning algorithm based on local smoothness is proposed (GIML). GIML algorithm first extracts the local smoothness structure of data set via local PCA. Then the optimal linear transformation, which transforms the local smoothness structure of new arrival sample’s neighborhood to its corresponded low-dimensional embedding coordinates, is computed. Finally the low-dimensinal embedding coordinates of new arrival samples are obtained by the optimal transformation. Extensive and systematic experiments are conducted on both artificial and real image data sets. Experimental results demonstrate that our GIML algotithm is an effective incremental manifold learning algorithm and outperforms other existing algirthms.

In real applications in pattern recognition and computer vison, high dimensional data always lie approximately on a low dimensional manifold. How to improve the performance of clustering algorithm on high dimensional data by using the manifold structure is a research hotspot in machine learning and data mining community. In this paper, a novel clustering algorithm called Nonlinear Discriminant K-means Clustering (NDisKmeans), which has taken the manifold structure of high dimensional into account, is proposed. By introducing the spectracl regularization technology, NDisKmeans first represents the desired low dimensional coordinates as linear combinations of smooth vectors predefined on the data manifold; then maximizes the ratio between inter-clusters scatter and total scatter to cluster the high dimensional data. A convergent iterative procedure is devised to solute the matrix of the combination coefficient and clustering assignment matrix. NDisKmeans overcomes the limilation of linear mapping of DisKmeans algorithm; therefore, it significantly improves the clustering performance. The systematic and extensive experiments on UCI and real world data sets have shown the effectiveness of the proposed NDisKmeans method.

Based on the characteristics of glass fragmentation images,a process was proposed.It first combined noise reduction,a traditional edge-detection operator,and threshold segmentation to produce an initial binary image partition.Then it applied the distance function to reconstruct and inverse it.Finally it used the watershed transformation based on chain code to obtain the whole segmentation image of fragmented glass.The final image can be used in testing safety glasses, for example to analyze and count the number of fragments.

A semiautomatic method for medical image segmentation based on active shape model was introduced. In order to improve the segmentation speed and precision, a semiautomatic method was used to model the training set, and a Gaussian Pyramid of images with different resolutions was generated so that multi-resolution image search could be performed. Experiments and analysis show that this method can be used to segment the lung region of medical images and the results are quite better.