對(duì)應(yīng)用統(tǒng)計(jì)方法評(píng)估核函數(shù)分類(lèi)能力的研究-資料下載頁(yè)
2025-06-27 14:03本頁(yè)面
【正文】 . and Qu . How to Add Transparency to Artificial Neural Networks? PR amp。 AI, 2007, 20(1): 7284 (in Chinese)（胡包鋼, 王泳, 楊雙紅, 曲寒冰. 如何增加人工神經(jīng)元網(wǎng)絡(luò)的透明度? 模式識(shí)別與人工智能, 2007: 20(1): 7284）[6] Yong Wang, Hu .. Study of the relationship between normalized information gain and accuracy、precision and recall. Beijing, 2007 Chinese Conference on Pattern Recognition (CCPR 2007), Beijing, Science Press, 2007: 2734 (in Chinese)(王泳, 胡包鋼. 歸一化信息增益準(zhǔn)則與準(zhǔn)確率、精確率、召回率的非線(xiàn)性關(guān)系研究. 2007年全國(guó)模式識(shí)別學(xué)術(shù)會(huì)議（CCPR2007）,北京,2007年12月,科學(xué)出版社, 2007: 2734)[7] Hu ., Yong Wang. Applications of Mutual Information Criteria in Classification Problems. Beijing, 2007 Chinese Conference on Pattern Recognition (CCPR 2007), Beijing, Science Press, 2007: 3545 (in Chinese)(胡包鋼, 王泳. 關(guān)于互信息學(xué)習(xí)準(zhǔn)則在分類(lèi)問(wèn)題中的應(yīng)用. 2007年全國(guó)模式識(shí)別學(xué)術(shù)會(huì)議（CCPR2007）,北京, 2007年12月, 科學(xué)出版社, 2007: 3545) BackgroundThis research is supported from National Natural Science Foundation of China, (“Nonlinearityvariationbased Study of Intelligent Systems”) and National Natural Science Foundation Outstanding Innovation Group Project, .The nonlinearvariation ability of functions refers to the ability of functions to approximate a cluster or multiclusters of nonlinear functions [1, 2]. Though some methods have been proposed to solve the problem from the aspects of “Nonlinear domain analysis” [3] and “the application of apriori knowledge” [4, 5], in view of practical applications, how to measure the “nonlinearvariation ability” quantitatively is still the most difficult and the key part of the research. How to choose the kernel functions is a special case of it, and up to date there is still not a practical framework to guide kernel selections. Analyzing the nonlinearvariation ability of functions may give us a new choice for kernel selection.In this paper, we applied statistical methods to study the problem of kernel selection quantitatively. Kfold crossvalidation is a monly used statistical method for kernel selection, while it is valid only if the independence between the data and the classifiers is guaranteed. In practical applications this premise condition is usually inaccessible. So by employing the corrected resample ttest – a newly proposed statistical method, we pare it with other two statistical methods – kfold crossvalidation and paired ttest on nine normally used kernels to measure their classification abilities. In addition, a new quantitative criterion of evaluating kernel classification performance based on information gain is proposed, which is proved to be the nonlinear function of traditional criteria and with wider application range [6, 7]. Benchmark tests show that the proposed quantitative statistical method is valid, and the information gain criterion is simple, stable. Furthermore, it can make up other criteria to a certain extent. Similar systematic studies on statistical methods of kernel selection are seldom reported in now.[1] Hu ., Mann G. K. I. and Gosine R. G. Control curve design for nonlinear (or fuzzy) proportional actions using splinebased functions. Automatica, 1998, 34(9): 11251133.[2] Hu ., Mann G. K. I. and Gosine R. G. A systematic study of fuzzy PID controllers – Functionbased evaluation approach. IEEE Trans. on Fuzzy Systems, 2001, 9(5): 699712.[3] Hu ., Xing . and Yang . Geometric Interpretation of Nonlinear Approximation Capability for Feedforward Neural Networks. ISNN 2004, Part I, LNCS 3137, 2004, 713.[4] Hu ., Qu . and Wang Y. Associating Neural Networks with Partially Known Relationships for Nonlinear Regressions. ICIC 2005, Part I, LNCS 3644, 2005, 737746.[5] Hu ., Wang Y., Yang . and Qu . How to Add Transparency to Artificial Neural Networks? PR amp。 AI, 2007, 20(1): 7284 (in Chinese)[6] Yong Wang, Hu .. Study of the relationship between normalized information gain and accuracy、precision and recall. Beijing, 2007 Chinese Conference on Pattern Recognition (CCPR 2007), Beijing, Science Press, 2007: 2734 (in Chinese)[7] Hu ., Yong Wang. Applications of Mutual Information Criteria in Classification Problems. Beijing, 2007 Chinese Conference on Pattern Recognition (CCPR 2007), Beijing, Science Press, 2007: 3545 (in Chinese)
外語(yǔ)相關(guān)推薦
鄂ICP備17016276號(hào)-1