This paper compares the sensitivity definitions of neural networks' output to input and weight *** on the essence of the sensitivity definitions, the authors classify these sensitivity definitions into 3 categorie...
详细信息
This paper compares the sensitivity definitions of neural networks' output to input and weight *** on the essence of the sensitivity definitions, the authors classify these sensitivity definitions into 3 categories: Noise-to- Signal Ratio, Geometrical property of derivative, Angle perturbation in Geometry *** characteristics of these 3 categories of sensitivity definition are discussed respectively, It is sensible to classify these sensitivity definitions based on the essence of them for researchers can find other new sensitivity definitions of neural networks.
作者:
Kwang-Hyun ParkZeungnam BienDivision of EE
Department of EECS Korea Advanced Institute of Science and Technology 373–1 Kusong-dong Yusong-gu Taejon 305–701 Korea. Zeungname Bien:received the B.S. degree in electronics engineering from Seoul National University
Seoul Korea in 1969 and the M.S. and Ph.D. degrees in electrical engineering from the University of Iowa Iowa City Iowa U.S.A. in 1972 and 1975 respectively. During 1976–1977 academic years he taught as assistant professor at the Department of Electrical Engineering University of Iowa. Then Dr. Bien joined Korea Advanced Institute of Science and Technology summer 1977 and is now Professor of Control Engineering at the Department of Electrical Engineering and Computer Science KAIST. Dr. Bien was the president of the Korea Fuzzy Logic and Intelligent Systems Society during 1990–1995 and also the general chair of IFSA World Congress 1993 and for FUZZ-IEEE99 respectively. He is currently co-Editor-in-Chief for International Journal of Fuzzy Systems (IJFS) Associate Editor for IEEE Transactions on Fuzzy Systems and a regional editor for the International Journal of Intelligent Automation and Soft Computing. He has been serving as Vice President for IFSA since 1997 and is now Chief Chairman of Institute of Electronics Engineers of Korea and Director of Humanfriendly Welfare Robot System Research Center. His current research interests include intelligent control methods with emphasis on fuzzy logic systems service robotics and rehabilitation engineering and large-scale industrial control systems. Kwang-Hyun Park:received the B.S.
M.S. and Ph.D. degrees in electrical engineering and computer science from KAIST Korea in 1994 19997 and 2001 respectively. He is now a researcher at Human-friendly Welfare Robot System Research Center. His research interests include learning control machine learning human-friendly interfaces and service robotics.
It has been found that some huge overshoot in the sense of sup-norm may be observed when typical iterative learning control (ILC) algorithms are applied to LTI systems, even though monotone convergence in the sense of...
详细信息
It has been found that some huge overshoot in the sense of sup-norm may be observed when typical iterative learning control (ILC) algorithms are applied to LTI systems, even though monotone convergence in the sense of λ-norm is guaranteed. In this paper, a new ILC algorithm with adjustment of learning interval is proposed to resolve such an undesirable phenomenon, and it is shown that the output error can be monotonically converged to zero in the sense of sup-norm when the proposed ILC algorithm is applied. A numerical example is given to show the effectiveness of the proposed algorithm.
We propose a novel bilingual topical admixture (BiTAM) formalism for word alignment in statistical machine translation. Under this formalism, the parallel sentence-pairs within a document-pair are assumed to constitut...
详细信息
A major challenge of computational biology is the inference of genetic regulatory networks and the identification of their topology from DNA microarray data. Recent results show that scale-free networks play an import...
详细信息
A major challenge of computational biology is the inference of genetic regulatory networks and the identification of their topology from DNA microarray data. Recent results show that scale-free networks play an important role in this context. These networks are characterized by a very small number of highly connected and relevant nodes, and by numerous poorly connected ones. In this paper, we experimentally assess the predictive power of the scale-free paradigm in a supervised learning framework. The hypotheses we intend to test in this framework are: (i) regulatory genes are effective predictors of the expression of the genes they regulate;(ii) a subset of regulatory genes may explain most of the variability of the measures. More precisely, we use the expression levels of a subset of regulatory genes, returned by feature selection, as input of a learningmachine which has to predict the expression levels of the target genes. We will show that (i) each gene can be predicted by a small subset of regulatory genes, and (ii) on a global scale, a small subset of regulatory genes, called the hubs, can have a non-negligible predictive power on all the target genes. Also, most of the regulatory genes returned by the application of this approach to Gasch et al. (2000) data were identified by Segal et al. (2003) and form a biologically coherent set of genes.
A novel and simple combination of inductive logic programming with kernel methods is presented. The kFOIL algorithm integrates the well-known inductive logic programming system FOIL with kernel methods. The feature sp...
详细信息
We apply Stochastic Meta-Descent (SMD), a stochastic gradient optimization method with gain vector adaptation, to the training of Conditional Random Fields (CRFs). On several large data sets, the resulting optimizer c...
详细信息
ISBN:
(纸本)1595933832
We apply Stochastic Meta-Descent (SMD), a stochastic gradient optimization method with gain vector adaptation, to the training of Conditional Random Fields (CRFs). On several large data sets, the resulting optimizer converges to the same quality of solution over an order of magnitude faster than limited-memory BFGS, the leading method reported to date. We report results for both exact and inexact inference techniques.
A human listener has the ability to follow a speaker's voice while others are speaking simultaneously;in particular, the listener can organize the time-frequency energy of the same speaker across time into a singl...
详细信息
Most work on preference learning has focused on pairwise preferences or rankings over individual items. In this paper, we present a method for learning preferences over sets of items. Our learning method takes as inpu...
详细信息
ISBN:
(纸本)1595933832
Most work on preference learning has focused on pairwise preferences or rankings over individual items. In this paper, we present a method for learning preferences over sets of items. Our learning method takes as input a collection of positive examples-that is, one or more sets that have been identified by a user as desirable. Kernel density estimation is used to estimate the value function for individual items, and the desired set diversity is estimated from the average set diversity observed in the collection. Since this is a new learning problem, we introduce a new evaluation methodology and evaluate the learning method on two data collections: synthetic blocks-world data and a new real-world music data collection that we have gathered.
This paper proposes a new definition of reduction in rough sets, which follows naturally from the concepts of the degree of similarity and the degree of inconsistency. The new definition is compared to the classical d...
详细信息
This paper proposes a new definition of reduction in rough sets, which follows naturally from the concepts of the degree of similarity and the degree of inconsistency. The new definition is compared to the classical definitions-the definition in algebra view and definition in information view,and is shown to be more general.
Semi-supervised learning algorithms have been successfully applied in many applications with scarce labeled data, by utilizing the unlabeled data. One important category is graph based semi-supervised learning algorit...
Semi-supervised learning algorithms have been successfully applied in many applications with scarce labeled data, by utilizing the unlabeled data. One important category is graph based semi-supervised learning algorithms, for which the performance depends considerably on the quality of the graph, or its hyperparameters. In this paper, we deal with the less explored problem of learning the graphs. We propose a graph learning method for the harmonic energy minimization method; this is done by minimizing the leave-one-out prediction error on labeled data points. We use a gradient based method and designed an efficient algorithm which significantly accelerates the calculation of the gradient by applying the matrix inversion lemma and using careful pre-computation. Experimental results show that the graph learning method is effective in improving the performance of the classification algorithm.
暂无评论