Marginal Fisher analysis (MFA) is a well-known linear dimensionality reduction method. However, MFA does not utilize the local diversity information of the training data, which will degrade its performance. In order t...
详细信息
Marginal Fisher analysis (MFA) is a well-known linear dimensionality reduction method. However, MFA does not utilize the local diversity information of the training data, which will degrade its performance. In order to enhance the discriminant power of MFA, this paper considers introducing local variation quantity to enlarge the distances between local neighborhood embeddings and proposes a flexible and efficient implementation of MFA (F-MFA) within the regularization framework. Therefore, the discriminant structure and diversity of data are preserved in low-dimensional subspace. Computationally, F-MFA is formulated as a trace differential optimization problem which can completely avoids the singularity problem as it exists in MFA. Further, an efficient algorithm is developed for implementing F-MFA via QR-decomposition. Experimental results on four face data sets demonstrate the effectiveness of our approach.
Named entity recognition is a traditional task in natural language processing. In particular, nested entity recognition receives extensive attention for the widespread existence of the nesting scenario. The latest res...
详细信息
Dictionary learning (DL) is powerful for representation learning, while it fails to capture the deep hierarchical information hidden in data. In this paper, we propose a new generalized end-to-end mulita-layer represe...
详细信息
ISBN:
(数字)9781728183169
ISBN:
(纸本)9781728183176
Dictionary learning (DL) is powerful for representation learning, while it fails to capture the deep hierarchical information hidden in data. In this paper, we propose a new generalized end-to-end mulita-layer representation learning architecture referred to as Multi-layer Dictionary Pair Learning Network (MDPL-net) for the deep sparse and hierarchical representation of images. To enable MDPL-net to conduct accurate classification, MDPL-net clearly integrates the skip connection end-to-end network and multi-layer deep sparse dictionary learning into a unified architecture. The representation learning module has several hidden DL blocks, where each hidden DL block has a dictionary pair learning (DPL) layer, a batch-norm layer and an activation function layer, and the DL blocks are connected in a feed-forward manner. To further improve the information flow and maintain the privileged features between different DL blocks, a novel skip dense connectivity pattern is deployed between hidden DL blocks, which can obtain more stable and discriminative features. The DPL layer jointly formulates the discriminative synthesis dictionary and analysis dictionary by minimizing reconstruction error within each batch over the feature maps from front layers. Extensive results on benchmark databases demonstrate the effectiveness of MDPL-net for discriminative representation and robust image classification.
Word segmentation is one necessary component for Asian language search engine, and probability dictionary is core component for statistical language model based word segmentation application. Manually marking is the t...
详细信息
With rapid urbanization and increasing traffic, distribution path planning has become an important challenging issue. Traditional vehicle distribution route planning mostly relies on static road information or histori...
详细信息
ISBN:
(纸本)9781728176505
With rapid urbanization and increasing traffic, distribution path planning has become an important challenging issue. Traditional vehicle distribution route planning mostly relies on static road information or historical traffic data for route planning. However, route planning based on such information may be unreasonable, because many uncertainty factors involve in the vehicle distribution process. Edge intelligence can help to handle such uncertainty factors, as it can offer smart services in close proximity to the Internet of Vehicles (IoV) environment with low latency and less cost. Therefore, in this paper, a model for dynamic distribution path planning with IoV and edge intelligence is constructed, and a corresponding planning strategy is proposed. The result of simulation experiments shows its effectiveness.
A multi-objective particle swarm optimization algorithm based on ε- dominance is proposed. The ε-dominance is applied to update the external set in order to obtain the Pareto set with better distribution, and the d...
详细信息
A multi-objective particle swarm optimization algorithm based on ε- dominance is proposed. The ε-dominance is applied to update the external set in order to obtain the Pareto set with better distribution, and the dynamic adjustment strategy, which made the algorithm achieving the search and approximation to the Pareto set, is adopted in the iterative process of the ε-Pareto solution set. Three benchmark cases were tested and the results show that this algorithm is much more efficient than the DNPSO.
This paper proposed a complex ontology evolution based method of extracting data, and also completely designed an extraction system, which consists of four important components: Resolver, Extractor, Consolidator and t...
详细信息
This paper proposed a complex ontology evolution based method of extracting data, and also completely designed an extraction system, which consists of four important components: Resolver, Extractor, Consolidator and the ontology construction components. The system gives priority to the construction of mini-ontology. When the user submits query keywords to the deep web query interface, the returned result will pass through the prior three components;after that, the final execution result will be returned to user in a unified form. This paper adopted an extraction method that is different from the general ontology extraction. More specifically, the ontology used in extraction here is dynamic evolution, which can adapt various data source better. Experimental results proved that this method could effectively extract the data in the query result pages.
Deep web could automatically produce web pages according to the query criteria of users. The report found most query result page store data information using table form. knowledge management, information retrieval, We...
详细信息
Deep web could automatically produce web pages according to the query criteria of users. The report found most query result page store data information using table form. knowledge management, information retrieval, Web mining, abstract extraction and so on were benefited from automatically understanding of table forms. The study web forms on the web information extraction and integration have important significance. This paper proposed a domain-specific ontology based strategy for integration tables, and this method could independent the structure of table. Experimental results confirm that this method could effectively improve the accuracy of integration.
A novel constant tamper-proofing software watermark technique based on H encryption function is presented. First we split the watermark into smaller pieces before encoding them using CLOC scheme. With the watermark pi...
详细信息
The complex mapping z larr z a is important in dynamical application, but it is not received as much attention in the literature as the mapping z larr z a +c, for there is not fractal structure by using escape time a...
详细信息
The complex mapping z larr z a is important in dynamical application, but it is not received as much attention in the literature as the mapping z larr z a +c, for there is not fractal structure by using escape time algorithm. This paper utilizes a new method named as distance ratio iteration method and discusses the iteration properties of the complex mapping z larr z a . The distance ratio iteration method can render the convergence region of the mapping, so the image has complex and self-similarity structure. This paper generates fractal image using distance ratio iteration method for various exponents of z larr z a and discusses their visual properties. There is rich detail fractal structure in the mapping z larr z a
暂无评论