Inferring the experimental variogram used in geostatistics commonly relies on the method-of-moments approach.Ideally,the available data-set used for calculating the experimental variogram should be drawn from a regula...Inferring the experimental variogram used in geostatistics commonly relies on the method-of-moments approach.Ideally,the available data-set used for calculating the experimental variogram should be drawn from a regular pattern.However,in practice the available data-set is typically sampled over a sparse pattern at irregularly spaced locations.Hence,some binning of the variogram cloud is required to obtain fair estimates of the experimental variogram.Grouping of the variogram data pairs as a result of conventional binning depends on parameters such as the main anisotropic directions and a regular definition of the lag vectors.These parameters are not based on the configuration of the variogram data pairs in the variogram cloud but on a segment of it that is arbitrarily predefined.Therefore,the conventional experimental variogram estimation approach is biased because of the strict configuration of the bins over the variogram cloud.In this paper,a new method of estimating experimental variograms is proposed.Lag vectors and their tolerances are decided in the proposed method from information in the variogram cloud：they are not influenced by any predefined directions.The proposed methodology is a well-founded,practicable and easy-to-automate approach for experimental variogram calculation using an irregularly sampled data-set.Comparison of results from the new method to those from the traditional approach is very encouraging.展开更多
In space feature quantization, the most important problem is designing an efficient and compact codebook. The hierarchical clustering approach successfully solves the problem of quantifying the feature space in a larg...In space feature quantization, the most important problem is designing an efficient and compact codebook. The hierarchical clustering approach successfully solves the problem of quantifying the feature space in a large vocabulary size. In this paper we propose to use a tree structure of hierarchical self-organizing-map （H-SOM） with the depth length equal to two and a high size of branch factors （50, 100, 200, 400, and 500）. Moreover, an incremental learning process of H-SOM is used to overcome the problem of the curse of the dimensionafity of space. The method is evaluated on three public datasets. Results exceed the current state-of-art retrieval performance on Kentucky and Oxford5k dataset. However, it is with less performance on the Holidays dataset. The experiment results indicate that the proposed tree structure shows significant improvement with a large number of branch factors.展开更多
文摘Inferring the experimental variogram used in geostatistics commonly relies on the method-of-moments approach.Ideally,the available data-set used for calculating the experimental variogram should be drawn from a regular pattern.However,in practice the available data-set is typically sampled over a sparse pattern at irregularly spaced locations.Hence,some binning of the variogram cloud is required to obtain fair estimates of the experimental variogram.Grouping of the variogram data pairs as a result of conventional binning depends on parameters such as the main anisotropic directions and a regular definition of the lag vectors.These parameters are not based on the configuration of the variogram data pairs in the variogram cloud but on a segment of it that is arbitrarily predefined.Therefore,the conventional experimental variogram estimation approach is biased because of the strict configuration of the bins over the variogram cloud.In this paper,a new method of estimating experimental variograms is proposed.Lag vectors and their tolerances are decided in the proposed method from information in the variogram cloud：they are not influenced by any predefined directions.The proposed methodology is a well-founded,practicable and easy-to-automate approach for experimental variogram calculation using an irregularly sampled data-set.Comparison of results from the new method to those from the traditional approach is very encouraging.
文摘In space feature quantization, the most important problem is designing an efficient and compact codebook. The hierarchical clustering approach successfully solves the problem of quantifying the feature space in a large vocabulary size. In this paper we propose to use a tree structure of hierarchical self-organizing-map （H-SOM） with the depth length equal to two and a high size of branch factors （50, 100, 200, 400, and 500）. Moreover, an incremental learning process of H-SOM is used to overcome the problem of the curse of the dimensionafity of space. The method is evaluated on three public datasets. Results exceed the current state-of-art retrieval performance on Kentucky and Oxford5k dataset. However, it is with less performance on the Holidays dataset. The experiment results indicate that the proposed tree structure shows significant improvement with a large number of branch factors.
