跨模态哈希检索以其较高的检索效率和较低的存储成本,在跨模态检索领域受到了广泛的关注.现有的跨模态哈希大多直接从多模态数据中学习哈希码,不能充分利用数据的语义信息,因此无法保证数据低维特征在模态间的分布一致性,解决这个问题...跨模态哈希检索以其较高的检索效率和较低的存储成本,在跨模态检索领域受到了广泛的关注.现有的跨模态哈希大多直接从多模态数据中学习哈希码,不能充分利用数据的语义信息,因此无法保证数据低维特征在模态间的分布一致性,解决这个问题的关键之一是要准确地度量多模态数据之间的相似度.为此,提出一种基于对抗投影学习的哈希(adversarial projection learning based Hashing for cross-modal retrieval,APLH)方法用于跨模态检索.利用对抗训练学习来自不同模态的低维特征,并保证低维特征在模态间的分布一致性.在此基础上,利用跨模态投影匹配约束(cross-modal projection matching,CMPM),最小化特征投影匹配分布和标签投影匹配分布之间的KL(Kullback-Leibler)散度,利用标签信息使数据低维特征之间的相似度结构与语义空间中的相似度结构趋于一致.此外,在哈希码学习阶段,引入加权余弦三元组损失进一步利用数据的语义信息;且为减小哈希码的量化损失,使用离散优化的方法优化哈希函数.在3个跨模态数据集MIRFlickr25K,NUS-WIDE,Wikipedia上,以不同码位计算mAP,且所提方法的mAP值均优于其他算法,验证了其在跨模态哈希检索上的优越性、鲁棒性以及CMPM的有效性.展开更多
When checking the ice shape calculation software,its accuracy is judged based on the proximity between the calculated ice shape and the typical test ice shape.Therefore,determining the typical test ice shape becomes t...When checking the ice shape calculation software,its accuracy is judged based on the proximity between the calculated ice shape and the typical test ice shape.Therefore,determining the typical test ice shape becomes the key task of the icing wind tunnel tests.In the icing wind tunnel test of the tail wing model of a large amphibious aircraft,in order to obtain accurate typical test ice shape,the Romer Absolute Scanner is used to obtain the 3D point cloud data of the ice shape on the tail wing model.Then,the batch-learning self-organizing map(BLSOM)neural network is used to obtain the 2D average ice shape along the model direction based on the 3D point cloud data of the ice shape,while its tolerance band is calculated using the probabilistic statistical method.The results show that the combination of 2D average ice shape and its tolerance band can represent the 3D characteristics of the test ice shape effectively,which can be used as the typical test ice shape for comparative analysis with the calculated ice shape.展开更多
A methodology is presented whereby a neural network is used to learn the inverse kinematic relationships of the position and orientation of a six joint manipulator. The arm solution for the orientation of a manipulato...A methodology is presented whereby a neural network is used to learn the inverse kinematic relationships of the position and orientation of a six joint manipulator. The arm solution for the orientation of a manipulator using a self organizing neural net is studied in this paper. A new training model of the self organizing neural network is proposed by thoroughly studying Martinetz, Ritter and Schulten′s self organizing neural network based on Kohonen′s self organizing mapping algorithm using a Widrow Hoff type error correction rule and closely combining the characters of the inverse kinematic relationship for a robot arm. The computer simulation results for a PUMA 560 robot show that the proposed method has a significant improvement over other methods documented in the references in self organizing capability and precision by training process.展开更多
A configurable ontology mapping approach based on different kinds of concept feature information is introduced in this paper. In this approach, ontology concept feature information is classified as five kinds, which r...A configurable ontology mapping approach based on different kinds of concept feature information is introduced in this paper. In this approach, ontology concept feature information is classified as five kinds, which respectively corresponds to five kinds of concept similarity computation methods. Many existing ontology mapping approaches have adopted the multi-feature reasoning, whereas not all feature information can be com- puted in the real ontology mapping and only fractional feature information needs to be selected in the mapping computation. Consequently a eonfigurable ontology mapping model is introduced, which is composed of CMT model, SMT model and related transformation model. Through the configurable model, users can conveniently select the most suitable features and configure the suitable weights. Simultaneously, a related 3-step ontology mapping approach is also introduced. Associated with the traditional name and instance learner-based ontology mapping approach, this approach is evaluated by an ontology mapping application example.展开更多
A new sub-pixel mapping method based on BP neural network is proposed in order to determine the spatial distribution of class components in each mixed pixel.The network was used to train a model that describes the rel...A new sub-pixel mapping method based on BP neural network is proposed in order to determine the spatial distribution of class components in each mixed pixel.The network was used to train a model that describes the relationship between spatial distribution of target components in mixed pixel and its neighboring information.Then the sub-pixel scaled target could be predicted by the trained model.In order to improve the performance of BP network,BP learning algorithm with momentum was employed.The experiments were conducted both on synthetic images and on hyperspectral imagery(HSI).The results prove that this method is capable of estimating land covers fairly accurately and has a great superiority over some other sub-pixel mapping methods in terms of computational complexity.展开更多
文摘跨模态哈希检索以其较高的检索效率和较低的存储成本,在跨模态检索领域受到了广泛的关注.现有的跨模态哈希大多直接从多模态数据中学习哈希码,不能充分利用数据的语义信息,因此无法保证数据低维特征在模态间的分布一致性,解决这个问题的关键之一是要准确地度量多模态数据之间的相似度.为此,提出一种基于对抗投影学习的哈希(adversarial projection learning based Hashing for cross-modal retrieval,APLH)方法用于跨模态检索.利用对抗训练学习来自不同模态的低维特征,并保证低维特征在模态间的分布一致性.在此基础上,利用跨模态投影匹配约束(cross-modal projection matching,CMPM),最小化特征投影匹配分布和标签投影匹配分布之间的KL(Kullback-Leibler)散度,利用标签信息使数据低维特征之间的相似度结构与语义空间中的相似度结构趋于一致.此外,在哈希码学习阶段,引入加权余弦三元组损失进一步利用数据的语义信息;且为减小哈希码的量化损失,使用离散优化的方法优化哈希函数.在3个跨模态数据集MIRFlickr25K,NUS-WIDE,Wikipedia上,以不同码位计算mAP,且所提方法的mAP值均优于其他算法,验证了其在跨模态哈希检索上的优越性、鲁棒性以及CMPM的有效性.
基金supported by the AG600 project of AVIC General Huanan Aircraft Industry Co.,Ltd.
文摘When checking the ice shape calculation software,its accuracy is judged based on the proximity between the calculated ice shape and the typical test ice shape.Therefore,determining the typical test ice shape becomes the key task of the icing wind tunnel tests.In the icing wind tunnel test of the tail wing model of a large amphibious aircraft,in order to obtain accurate typical test ice shape,the Romer Absolute Scanner is used to obtain the 3D point cloud data of the ice shape on the tail wing model.Then,the batch-learning self-organizing map(BLSOM)neural network is used to obtain the 2D average ice shape along the model direction based on the 3D point cloud data of the ice shape,while its tolerance band is calculated using the probabilistic statistical method.The results show that the combination of 2D average ice shape and its tolerance band can represent the 3D characteristics of the test ice shape effectively,which can be used as the typical test ice shape for comparative analysis with the calculated ice shape.
文摘A methodology is presented whereby a neural network is used to learn the inverse kinematic relationships of the position and orientation of a six joint manipulator. The arm solution for the orientation of a manipulator using a self organizing neural net is studied in this paper. A new training model of the self organizing neural network is proposed by thoroughly studying Martinetz, Ritter and Schulten′s self organizing neural network based on Kohonen′s self organizing mapping algorithm using a Widrow Hoff type error correction rule and closely combining the characters of the inverse kinematic relationship for a robot arm. The computer simulation results for a PUMA 560 robot show that the proposed method has a significant improvement over other methods documented in the references in self organizing capability and precision by training process.
基金Sponsored by the 973 Natural Key Basis Research and Development Plan (Grant No.973: 2003CB316905)the National Natural Science Foundationof China (Grant No.60374071)
文摘A configurable ontology mapping approach based on different kinds of concept feature information is introduced in this paper. In this approach, ontology concept feature information is classified as five kinds, which respectively corresponds to five kinds of concept similarity computation methods. Many existing ontology mapping approaches have adopted the multi-feature reasoning, whereas not all feature information can be com- puted in the real ontology mapping and only fractional feature information needs to be selected in the mapping computation. Consequently a eonfigurable ontology mapping model is introduced, which is composed of CMT model, SMT model and related transformation model. Through the configurable model, users can conveniently select the most suitable features and configure the suitable weights. Simultaneously, a related 3-step ontology mapping approach is also introduced. Associated with the traditional name and instance learner-based ontology mapping approach, this approach is evaluated by an ontology mapping application example.
基金Sponsored by the National Natural Science Foundation of China(Grant No. 60272073, 60402025 and 60802059)by Foundation for the Doctoral Program of Higher Education of China (Grant No. 200802171003)
文摘A new sub-pixel mapping method based on BP neural network is proposed in order to determine the spatial distribution of class components in each mixed pixel.The network was used to train a model that describes the relationship between spatial distribution of target components in mixed pixel and its neighboring information.Then the sub-pixel scaled target could be predicted by the trained model.In order to improve the performance of BP network,BP learning algorithm with momentum was employed.The experiments were conducted both on synthetic images and on hyperspectral imagery(HSI).The results prove that this method is capable of estimating land covers fairly accurately and has a great superiority over some other sub-pixel mapping methods in terms of computational complexity.