1,748,011 research outputs found
Spirito didattico e sguardo artistico : etica e poetica di Stifter nei saggi letterari e nelle recensioni pittoriche e teatrali
No full text
Hybrid Metaheuristics: An Introduction
No full textIn many real life settings, high quality solutions to hard optimization problems such as flight scheduling or load balancing in telecommunication networks are required in a short amount of time. Due to the practical importance of optimization problems for industry and science, many algorithms to tackle them have been developed. One important class of such algorithms are metaheuristics. The field of metaheuristic research has enjoyed a considerable popularity in the last decades. In this introductory chapter we first provide a general overview on metaheuristics. Then we turn towards a new and highly successful branch of metaheuristic research, namely the hybridization of metaheuristics with algorithmic components originating from other techniques for optimization. The chapter ends with an outline of the remaining book chapters
Methods for Designing Multiple Classifier Systems
No full textIn the field of pattern recognition, multiple classifier systems based on the combination of outputs of a set of different classifiers have been proposed as a method for the development of high performance classification systems. In this paper, the problem of design of multiple classifier system is discussed. Six design methods based on the so-called “overproduce and choose“ paradigm are described and compared by experiments. Although these design methods exhibited some interesting features, they do not guarantee to design the optimal multiple classifier system for the classification task at hand. Accordingly, the main conclusion of this paper is that the problem of the optimal MCS design still remains ope
Statistical and neural classifiers: an integrated approach to design (Advances in Pattern Recognition Series)
No full textDynamic Classifier Selection
No full textAt present, the usual operation mechanism of multiple classifier systems is the combination of classifier outputs. Recently, some researchers have pointed out the potentialities of “dynamic classifier selection’ as an alternative operation mechanism. However, such potentialities have been motivated so far by experimental results and qualitative arguments. This paper is aimed to provide a theoretical framework for dynamic classifier selection and to define the assumptions under which it can be expected to improve the accuracy of the individual classifiers. To this end, dynamic classifier selection is placed in the general framework of statistical decision theory and it is shown that, under some assumptions, the optimal Bayes classifier can be obtained by selecting non-optimal classifiers. Two classifier selection methods that derive from the proposed framework are described. The experimental results obtained in the classification of remote-sensing images and comparisons among different combination methods are reported
Comparison and Combination of Adaptive Query Shifting and Feature Relevance Learning for Content-Based Image Retrieval
No full textBrowse Conference Publications > Image Analysis and Processing ...
Comparison and combination of adaptive query shifting and feature relevance learning for content-based image retrieval
This paper appears in:
Image Analysis and Processing, 2001. Proceedings. 11th International Conference on
Date of Conference: 26-28 Sep 2001
Author(s): Giacinto, G.
Dept. of Electr. & Electron. Eng., Cagliari Univ.
Roli, F. ; Fumera, G.
Page(s): 422 - 427
Product Type: Conference Publications
ABSTRACT
Despite the efforts to reduce the semantic gap between user perception of similarity and feature-based representation of images, user interaction is essential to improve retrieval performance in content-based image retrieval. To this end a number of relevance feedback mechanisms are currently adopted to refine image queries. They are aimed either to locally modify the feature space or to shift the query point towards more promising regions of the feature space. A novel adaptive query shifting mechanism is proposed to improve retrieval performance beyond that provided by other relevance feedback mechanisms. In addition we discuss the extent to which query shifting may provide better performance than feature weighting and provide experimental results on the complementarity of the two approaches. Finally, some combinational approaches are proposed to exploit such complementarities
Hybrid Metaheuristics: Preface to the proceedings of HM2005
No full textCombinatorial optimization attracted many researchers since more than three decades.
Plenty of classical hard problems have been tackled successfully with metaheuristic approaches. Several thereof are currently considered state-of-the-art methods for such problems. However, for many years the main focus of research was on the application of single metaheuristics to given problems. A tendency to compare different metaheuristics against each other could be observed, and sometimes this competition led to thinking in stereotypes in the
research communities.
In recent years, it has become evident that the concentration on a sole metaheuristic is rather restrictive, when focusing on the improvement of heuristic techniques to tackle both academic and practical optimization problems. A skilled combination of concepts stemming from different metaheuristics can provide a more efficient behavior and a higher flexibility. Also the hybridization of metaheuristics with other techniques known from classical artificial intelligence areas can be very fruitful. Further, the incorporation of typical operations research techniques can be very beneficial. Combinations of metaheuristic components with components from other metaheuristics or optimization strategies from artificial intelligence or operations research are called hybrid metaheuristics.
The design and implementation of hybrid metaheuristics rises problems going beyond questions about the composition of a single metaheuristic. The proper interaction of
different algorithm components must usually be based on a careful analysis of the single components. Choice and tuning of parameters is more important for the quality of the algorithms than before. Different concepts of interaction at low-level and at high-level are studied. As a result, the design of experiments and the proper statistical evaluation are in a more exposed position than before.
We believe that the combination of elements coming from different metaheuristics, and from classical methods from both artificial intelligence and operations research, bears great chances to become one of the main tracks of research in applied artificial intelligence. It seems to be a promising and rewarding alternative to the still existing mutual contempt between the fields of exact methods and approximate techniques, and also to the competition between the different schools of metaheuristics, which sometimes focused more on a proof of concept than on good general results.
Still, we have to realize that research on hybrid metaheuristics is in main parts based on experimental methods, thus being probably more related to natural sciences than to computer science. It can be stated that both the design and the evaluation of experiments have still not reached the standard as they have in physics or chemistry
for example. The validity of analyses of experimental work on algorithms is a key aspect in hybrid metaheuristics, and the attention of researchers to this aspect seems to be important for the future of the field
Self-loops favour diversification and asymmetric transitions between attractors in boolean network models
No full textThe process of cell differentiation manifests properties such as non-uniform robustness and asymmetric transitions among cell types. In this paper we adopt Boolean networks to model cellular differentiation, where attractors (or set of attractors) in the network landscape epitomise cell types. Since changes in network topology and functions strongly impact attractor landscape characteristics, in this paper we study how self-loops influence diversified robustness and asymmetry of transitions. The purpose of this study is to identify the best configuration for a network owning these properties. Our results show that a moderate amount of self-loops make random Boolean networks more suitable to reproduce differentiation phenomena. This is a further evidence that self-loops play an important role in genetic regulatory networks
- …
