Traditional methods for selecting models in experimental data analysis are susceptible to researcher bias, hindering exploration of alternative explanations and potentially leading to overfitting. The Finite Informati...Traditional methods for selecting models in experimental data analysis are susceptible to researcher bias, hindering exploration of alternative explanations and potentially leading to overfitting. The Finite Information Quantity (FIQ) approach offers a novel solution by acknowledging the inherent limitations in information processing capacity of physical systems. This framework facilitates the development of objective criteria for model selection (comparative uncertainty) and paves the way for a more comprehensive understanding of phenomena through exploring diverse explanations. This work presents a detailed comparison of the FIQ approach with ten established model selection methods, highlighting the advantages and limitations of each. We demonstrate the potential of FIQ to enhance the objectivity and robustness of scientific inquiry through three practical examples: selecting appropriate models for measuring fundamental constants, sound velocity, and underwater electrical discharges. Further research is warranted to explore the full applicability of FIQ across various scientific disciplines.展开更多
In variational problem, the selection of functional weighting factors (FWF) is one of the key points for discussing many relevant studies. To overcome arbitrariness and subjectivity of the empirical selecting methods ...In variational problem, the selection of functional weighting factors (FWF) is one of the key points for discussing many relevant studies. To overcome arbitrariness and subjectivity of the empirical selecting methods used widely at present, this paper tries to put forward an optimal objective selecting method of FWF. The focus of the study is on the weighting factors optimal selection in the variation retrieval single-Doppler radar wind field with the simple adjoint models. Weighting factors in the meaning of minimal variance are calculated out with the matrix theory and the finite difference method of partial differential equation. Experiments show that the result is more objective comparing with the factors obtained with the empirical method.展开更多
Feature selection is a key task in statistical pattern recognition. Most feature selection algorithms have been proposed based on specific objective functions which are usually intuitively reasonable but can sometimes...Feature selection is a key task in statistical pattern recognition. Most feature selection algorithms have been proposed based on specific objective functions which are usually intuitively reasonable but can sometimes be far from the more basic objectives of the feature selection. This paper describes how to select features such that the basic objectives, e.g., classification or clustering accuracies, can be optimized in a more direct way. The analysis requires that the contribution of each feature to the evaluation metrics can be quantitatively described by some score function. Motivated by the conditional independence structure in probabilistic distributions, the analysis uses a leave-one-out feature selection algorithm which provides an approximate solution. The leave-one- out algorithm improves the conventional greedy backward elimination algorithm by preserving more interactions among features in the selection process, so that the various feature selection objectives can be optimized in a unified way. Experiments on six real-world datasets with different feature evaluation metrics have shown that this algorithm outperforms popular feature selection algorithms in most situations.展开更多
文摘Traditional methods for selecting models in experimental data analysis are susceptible to researcher bias, hindering exploration of alternative explanations and potentially leading to overfitting. The Finite Information Quantity (FIQ) approach offers a novel solution by acknowledging the inherent limitations in information processing capacity of physical systems. This framework facilitates the development of objective criteria for model selection (comparative uncertainty) and paves the way for a more comprehensive understanding of phenomena through exploring diverse explanations. This work presents a detailed comparison of the FIQ approach with ten established model selection methods, highlighting the advantages and limitations of each. We demonstrate the potential of FIQ to enhance the objectivity and robustness of scientific inquiry through three practical examples: selecting appropriate models for measuring fundamental constants, sound velocity, and underwater electrical discharges. Further research is warranted to explore the full applicability of FIQ across various scientific disciplines.
文摘In variational problem, the selection of functional weighting factors (FWF) is one of the key points for discussing many relevant studies. To overcome arbitrariness and subjectivity of the empirical selecting methods used widely at present, this paper tries to put forward an optimal objective selecting method of FWF. The focus of the study is on the weighting factors optimal selection in the variation retrieval single-Doppler radar wind field with the simple adjoint models. Weighting factors in the meaning of minimal variance are calculated out with the matrix theory and the finite difference method of partial differential equation. Experiments show that the result is more objective comparing with the factors obtained with the empirical method.
基金National Natural Science Foundation of China(Nos.61071131 and 61271388)Beijing Natural Science Foundation(No.4122040)+1 种基金Research Project of Tsinghua University(No.2012Z01011)Doctoral Fund of the Ministry of Education of China(No.20120002110036)
文摘Feature selection is a key task in statistical pattern recognition. Most feature selection algorithms have been proposed based on specific objective functions which are usually intuitively reasonable but can sometimes be far from the more basic objectives of the feature selection. This paper describes how to select features such that the basic objectives, e.g., classification or clustering accuracies, can be optimized in a more direct way. The analysis requires that the contribution of each feature to the evaluation metrics can be quantitatively described by some score function. Motivated by the conditional independence structure in probabilistic distributions, the analysis uses a leave-one-out feature selection algorithm which provides an approximate solution. The leave-one- out algorithm improves the conventional greedy backward elimination algorithm by preserving more interactions among features in the selection process, so that the various feature selection objectives can be optimized in a unified way. Experiments on six real-world datasets with different feature evaluation metrics have shown that this algorithm outperforms popular feature selection algorithms in most situations.
