检索结果-内蒙古大学图书馆

Quantizability and learning complexity in multilayer neural networks

IEEE TRANSACTIONS ON SYSTEMS MAN AND CYBERNETICS PART C-APPLICATIONS AND REVIEWS 1998年第2期28卷 295-300页

作者： Fu, LM Univ Florida Dept Comp & Informat Sci Gainesville FL 32611 USA

The relationship between quantizability and learning complexity in multilayer neural networks is examined. In a special neural network architecture that calculates node activations according to the certainty factor (CF) model of expert systems, the analysis based upon quantizability leads to lower and also better estimates for generalization dimensionality and sample complexity than those suggested by the multilayer perceptron model. This analysis is further supported by empirical simulation results.

关键词： generalization learning complexity machine learning neural network quantizability sample complexity

来源：评论

学校读者我要写书评

暂无评论

Bounds for the computational power and learning complexity of analog neural nets

引用

SIAM JOURNAL ON COMPUTING 1997年第3期26卷 708-732页

作者： Maass, W Institute for Theoretical Computer Science Technische Universitat Graz

It is shown that high-order feedforward neural nets of constant depth with piecewise-polynomial activation functions and arbitrary real weights can be simulated for Boolean inputs and outputs by neural nets of a somewhat larger size and depth with Heaviside gates and weights from {-1, 0, 1}. This provides the first known upper bound for the computational power of the former type of neural nets. It is also shown that in the case of first-order nets with piecewise-linear activation functions one can replace arbitrary real weights by rational numbers with polynomially many bits without changing the Boolean function that is computed by the neural net. In order to prove these results, we introduce two new methods for reducing nonlinear problems about weights in multilayer neural nets to linear problems for a transformed set of parameters. These transformed parameters can be interpreted as weights in a somewhat larger neural net. As another application of our new proof technique we show that neural nets with piecewise-polynomial activation functions and a constant number of analog inputs are probably approximately correct (PAC) learnable (in Valiant's model for PAC learning [Comm. Assoc. Comput. Mach., 27 (1984), pp. 1134-1142]).

关键词： neural networks analog computing threshold circuits circuit complexity learning complexity

来源：评论

学校读者我要写书评

暂无评论

Measuring learning complexity with Criteria Epitomizers

Measuring Learning Complexity with Criteria Epitomizers

引用

28th International Symposium on Theoretical Aspects of Computer Science (SATCS)

作者： Case, John Koetzing, Timo Univ Delaware Dept Comp & Informat Sci Newark DE 19716 USA Max Planck Inst Informat D-66123 Saarbrucken Germany

ISBN: (纸本)9783939897255

In prior papers, beginning with the seminal work by Freivalds et al. 1995, the notion of intrinsic complexity is used to analyze the learning complexity of sets of functions in a Gold-style learning setting. Herein are pointed out some weaknesses of this notion. Offered is an alternative based on epitomizing sets of functions - sets, which are learnable under a given learning criterion, but not under other criteria which are not at least as powerful. To capture the idea of epitomizing sets, new reducibility notions are given based on robust learning (closure of learning under certain classes of operators). Various degrees of epitomizing sets are characterized as the sets complete with respect to corresponding reducibility notions! These characterizations also provide an easy method for showing sets to be epitomizers, and they are, then, employed to prove several sets to be epitomizing. Furthermore, a scheme is provided to generate easily very strong epitomizers for a multitude of learning criteria. These strong epitomizers are so-called self-learning sets, previously applied by Case & Kotzing, 2010. These strong epitomizers can be generated and employed in a myriad of settings to witness the strict separation in learning power between the criteria so epitomized and other not as powerful criteria!

关键词： Algorithmic learning Theory learning complexity Robustness in learning

来源：评论

学校读者我要写书评

暂无评论

The complexity of learning concept classes with polynomial general dimension

引用

THEORETICAL COMPUTER SCIENCE 2006年第1期350卷 49-62页

作者： Köbler, J Lindner, W Humboldt Univ Inst Informat D-10099 Berlin Germany Univ Ulm Theoret Informat Abt D-89069 Ulm Germany

The general dimension is a combinatorial measure that characterizes the number of queries needed to learn a concept class. We use this notion to show that any p-evaluatable concept class with polynomial query complexity can be learned in polynomial time with the help of an oracle in the polynomial hierarchy, where the complexity of the required oracle depends on the query-types used by the learning algorithm. In particular, we show that for subset and superset queries an oracle in Sigma(P)(3) suffices. Since the concept class of DNF formulas has polynomial query complexity with respect to subset and superset queries with DNF formulas as hypotheses, it follows that DNF formulas are properly learnable in polynomial time with subset and superset queries and the help of an oracle in Sigma(P)(3). We also show that the required oracle in our main theorem cannot be replaced by an oracle in a lower level of the polynomial-time hierarchy, unless the hierarchy collapses. (c) 2005 Elsevier B.V. All rights reserved.

关键词： query learning learning complexity learning DNF formulas polynomial-time hierarchy

来源：评论

学校读者我要写书评

暂无评论

The complexity of learning concept classes with polynomial general dimension

The complexity of learning concept classes with polynomial g...

引用

13th Annual International Conference on Algorithmic learning Theory

作者： Köbler, J Lindner, W Humboldt Univ Inst Informat D-10099 Berlin Germany Univ Ulm Theoret Informat Abt D-89069 Ulm Germany

ISBN: (纸本)3540001700

关键词： query learning learning complexity learning DNF formulas polynomial-time hierarchy

来源：评论

学校读者我要写书评

暂无评论

Iterative generation of higher-order nets in polynomial time using linear programming

引用

IEEE TRANSACTIONS ON NEURAL NETWORKS 1997年第2期8卷 402-412页

作者： Roy, A Mukhopadhyay, S Laboratory of Chromatography DEPg.Fac.Quimica Universidad Nacional Autonoma de Mexico Circuito interior Cd Universitaria/CP 04510 Mexico D.F.Mexico

This paper presents an algorithm for constructing and training a class of higher-order perceptrons for classification problems. The method uses linear programming models to construct and train the net, Its polynomial time complexity is proven and computational results are provided for several well-known problems, In all cases, very small nets were created compared to those reported in other computational studies.

关键词： designing neural networks feedforward nets higher-order networks learning complexity linear programming polynomial time complexity multilayer perceptrons

来源：评论

学校读者我要写书评

暂无评论

Neural-Network-Based Constrained Output-Feedback Control for MEMS Gyroscopes Considering Scarce Transmission Bandwidth

引用

IEEE TRANSACTIONS ON CYBERNETICS 2022年第11期52卷 12351-12363页

作者： Shao, Xingling Shi, Yi North Univ China Minist Educ Key Lab Instrumentat Sci & Dynam Measurement Taiyuan 030051 Peoples R China North Univ China Sch Instrument & Elect Taiyuan 030051 Peoples R China

In this article, a neural-network-based constrained output-feedback control is considered for microelectromechanical system (MEMS) gyroscopes subject to scarce transmission bandwidth and lumped disturbances resulting from model uncertainties, dynamic coupling, and environmental disturbances. First, a hybrid quantizer capable of achieving an adjustable communication rate and quantization density is proposed to convert continuous control signals into discrete values, allowing for reduced chattering behavior even when control actions vary within large regions and enhanced tracking accuracy can be ensured. Subsequently, by applying two types of nonlinear mapping, all state variables of MEMS gyroscopes are restrained within the predefined time-varying asymmetric functions without imposing stringent feasibility conditions on virtual control laws. Furthermore, an echo-state network-based minimal learning parameter neural observer is developed to simultaneously recover the unmeasurable velocity-state variables, matched as well as unmatched disturbances in constrained MEMS gyroscopes dynamics, enabling an output-feedback control solution with a decreased online learning complexity. It is shown via the Lyapunov stability and nonsmooth analysis that all signals in the closed-loop system remain ultimately uniformly bounded even with discontinuous control actions. Comparison simulations are produced to certify the effectiveness of the presented controller.

关键词： Micromechanical devices Gyroscopes Quantization (signal) Bandwidth Artificial neural networks Velocity measurement Nonlinear systems Hybrid quantizer (HQ) learning complexity microelectromechanical system (MEMS) gyroscopes neural adaptive output feedback

来源：评论

学校读者我要写书评

暂无评论

On teaching sets of k-threshold functions

引用

INFORMATION AND COMPUTATION 2016年 251卷 301-313页

作者： Zamaraeva, Elena Nizhnii Novgorod State Univ Gagarin Ave 23 Nizhnii Novgorod 603600 Russia

In this paper we study teaching sets of k-threshold functions, i.e. functions that can be represented as a conjunction of k threshold functions. We reveal a connection between essential points of k threshold functions and essential points of the corresponding k-threshold function. We show that in two-dimensional case the number of minimal teaching sets of a 2-threshold function can grow as Omega(n(2)). We also consider the class of polytopes with vertices in the d-dimensional cube. Each polytope from this class can be defined by a k-threshold function for some k. We prove that such a polytope has a unique minimal teaching set which is equal to the set of its essential points. For d = 2 we describe structure of the minimal teaching set of a polytope and show that its cardinality is either Theta(n(2)) or O(n) and depends on the perimeter and the minimum angle of the polytope. it) (C) 2016 Elsevier Inc. All rights reserved.

关键词： Machine learning Threshold function Essential point Teaching set learning complexity k-threshold function

来源：评论

学校读者我要写书评

暂无评论

Accuracy-based learning Classifier Systems:: Models, analysis and applications to classification tasks

引用

EVOLUTIONARY COMPUTATION 2003年第3期11卷 209-238页

作者： Bernadó-Mansilla, E Garrell-Guiu, JM Ramon Llull Univ Barcelona 08022 Spain

Recently, learning Classifier Systems (LCS) and particularly XCS have arisen as promising methods for classification tasks and data ruining. This paper investigates two models of accuracy-based learning classifier systems oil different types of classification problems. Departing from XCS, we analyze the evolution of a complete action map as a knowledge representation. We propose an alternative, UCS, which evolves a best action map more efficiently. We also investigate how the fitness pressure guides the search towards accurate classifiers. While XCS bases fitness oil a reinforcement learning scheme, UCS defines fitness from a supervised learning scheme. We find significant differences in how the fitness pressure leads towards accuracy, and suggest the use of a supervised approach specially for multi-class problems and problems with unbalanced classes. We also investigate the complexity factors which arise ill each type of accuracy-based LCS. We provide a model on the learning complexity of LCS which is based oil the representative examples given to the system. The results and observations are also extended to a set of real world classification problems, where accuracy-based LCS are shown to perform competitively with respect to other learning algorithms. The work presents an extended analysis of accuracy-based LCS, gives insight into the Understanding of the LCS dynamics, and suggests open issues for further improvement of LCS on classification tasks.

关键词： learning classifier systems accuracy-based fitness knowledge representation learning complexity generalization data mining

来源：评论

学校读者我要写书评

暂无评论

complexity of hyperconcepts

引用

THEORETICAL COMPUTER SCIENCE 2006年第1期363卷 2-10页

作者： Ratsaby, Joel Ben Gurion Univ Negev IL-84105 Beer Sheva Israel

In machine-learning, maximizing the sample margin can reduce the teaming generalization error. Samples on which the target function has a large margin (gamma) convey more information since they yield more accurate hypotheses. Let X be a finite domain and S denote the set of all samples S subset of X of fixed cardinality in. Let H be a class of hypotheses h on X. A hyperconcept h' is defined as an indicator function for a set A subset of S of all samples on which the corresponding hypothesis h has a margin of at least gamma. An estimate on the complexity of the class H' of hyperconcepts h' is obtained with explicit dependence on gamma, the pseudo-dimension of H and m. (C) 2006 Elsevier B.V All rights reserved.

关键词： sample-dependent error-bounds large-margin samples learning complexity pseudo-dimension

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：