Adaptive resolution min-max classifiers

Abstract

A high automation degree is one of the most important features of data driven modeling tools and it should be taken into consideration in classification systems design. In this regard, constructive training algorithms are essential to improve the automation degree of a modeling system. Among neuro-fuzzy classifiers, Simpson's (1992) min-max networks have the advantage of being trained in a constructive way. The use of the hyperbox, as a frame on which different membership functions can be tailored, makes the min-max model a flexible tool. However, the original training algorithm evidences some serious drawbacks, together with a low automation degree. In order to overcome these inconveniences, in this paper two new learning algorithms for fuzzy min-max neural classifiers are proposed: the adaptive resolution classifier (ARC) and its pruning version (PARC). ARC/PARC generates a regularized min-max network by a succession of hyperbox cuts. The generalization capability of ARC/PARC technique mostly depends on the adopted cutting strategy. By using a recursive cutting procedure (R-ARC and R-PARC) it is possible to obtain better results. ARC, PARC, R-ARC, and R-PARC are characterized by a high automation degree and allow to achieve networks with a remarkable generalization capability. Their performances are evaluated through a set of toy problems and real data benchmarks. The paper also proposes a suitable index that can be used for the sensitivity analysis of the classification systems under consideration.

Granular Computing Techniques for Classification and Semantic Characterization of Structured Data

“…At this stage, a generic classifier conceived to work in a vector space can easily deal with the classification task. We decided to use a k-NN classifier, but different methods could be adopted, for example, several types of single layer feed-forward networks (e.g., neurofuzzy networks or random vector functional-links [34,38]). As in the case of the Gk-NN classifier, the parameter k is chosen a-priori by the user.…”

Mạng nơron min-max mờ tăng trưởng

“…Đến nay, có một số nghiên cứu đề xuất cải tiến hiệu suất thuật toán học của FMNN. Rizzi đã cải tiến FMNN của Simpson bằng cách áp dụng các kỹ thuật phân loại thích ứng [17], cắt tỉa [18], khái quát PARC (Pruning Adaptive Resolution Classifier) [19] sử dụng kỹ thuật cài đặt đệ quy. Tuy nhiên, các giải pháp này có chi phí tính toán lớn, do hạn chế của việc sử dụng đệ quy [18].…”

The design of granular classifiers: A study in the synergy of interval calculus and fuzzy sets in pattern recognition

“…Abe et al [3][4][5] presented an efficient method for extracting rules directly from a series of activation hyperboxes, which capture the existence region of data for a given class and inhibition hyperboxes, which inhibit the existence of data of that class. Rizzi et al [6,7] proposed an adaptive resolution classifier (ARC) and its pruned version (PARC) in order to enhance the constructs introduced by Simpson. ARC/PARC generates a regularized min-max network by a series of hyperbox cuts.…”

“…Each classifier comes with its geometry and this predominantly determines its capabilities [1][2][3][4][5][6][7][8][9]. While linear classifiers (built on a basis of some hyperplanes) and nonlinear classifiers (such as neural networks) are two popular alternatives, there is another point of view at the development of the classifiers that dwells on the concept of information granules.…”

“…PARC [7] 1.66 4 4. OMM [7] 3. to. This situation is very much reflective of the existing overlap between the classes in such regions.…”