S.M. WEISS and I. KAPOULEAS, 1989. An Empirical Comparison of Pattern Recognition, Neural Nets, and Machine Learning Classification Methods, in Proceedings of the Eleventh International Joint Conference on Artificial Intelligence, Morgan Kaufmann, San Mateo, CA, 688 -- 693.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....soybean diseases, chess end games, audiological disorders, and the Nettalk data set. Their conclusion wfis that the accuracy of classifying new examples was about the same for all the three systems, but the neural net performed better than ID3 when there was noise in the data. Weiss and Kapouleas [5] compared ID3, predictive value maximization, neural net using backpropagation, and a few statistical methods. They found that the statistical classifiers performed consistently better in terms of accuracy in classifying testing examples. Dietterich, Hlld, and Bakiri [6] compared ID3 with a ....

Weiss, S.M. and Kapouleas, I. 1989. An Empirical Comparison of Pattern Recognition, Neural Nets, and Machine Learning Classification Methods. In Proceedings of 1JCAI-89. Detroit, MI: Morgan Kaufmann.

....by more than one node. This makes a big network totally ununderstandable. Symbolic learning techniques produce more understandable outputs but they are not as good as connectionist learning techniques in generalization [Atlas et al. 1990, Shavlik et al. 1991, Fisher and McKusick, 1989, Weiss and Kapouleas, 1990] Some of the symbolic learning algorithms can create If Then rules for representing the knowledge they acquire from the dataset. If Then rules are important not only because they are more understandable but also because they can be used in creation of expert systems. Knowledge acquisition is ....

Weiss, S. M. and Kapouleas, I. (1990). An empirical comparison of pattern recognition, neural nets, and machine learning classi cation 146 methods. In Shavlik, J. W. and Dietterich, T. G., editors, Readings in Machine Learning. Morgan Kaufman, San Mateo, CA.

....implementing them listed below. 3.1 Decision Trees Decision trees are perhaps the most widely studied inductive learning models in the machine learning community. The literature abounds with papers proposing new models or variations of existing models and case studies using decision trees ([14, 21, 22, 25, 30, 34, 22 40, 43, 49, 50, 51, 53, 89, 93, 98, 99, 100, 101, 102, 104, 105, 106, 107, 109, 110, 111, 112, 113, 114, 118, 120, 123, 126, 129, 130, 131, 133, 134, 136]) For this case study, we use decision tree software from Quinlan and Buntine. Quinlan introduces decision trees and illustrates the use of his C4.5 software for decision trees (c4.5tree) and production rules derived therefrom (c4.5rule) in [105] Several decision tree algorithms (cart, id3, c4, ....

S.M. Weiss and I. Kapouleas (1990). An empirical comparison of pattern recognition, neural nets, and machine learning classification methods. Proceedings of the Eleventh International Joint Conference on Artificial Intelligence. Morgan Kaufmann, San Mateo, CA.781-787.

.... statistics [ Cherkaoui and Cleroux, 1991, Remme et al. 1980 ] and within neural networks [ Huang et al. 1991, Fahlman, 1991, Xu et al. 1991, Ersoy and Hong, 1991 ] There have been fewer, 3 but still many, inter subject studies involving algorithms from two or more of these fields, e.g. Weiss and Kapouleas, 1989, Weiss and Kulikowski, 1991 ] Ripley, 1992 ] Mooney et al. 1989, Shavlik et al. 1991 ] Fisher and McKusick, 1989 ] Thrun et al. 1991 ] Kirkwood et al. 1989 ] Tsaptsinos et al. 1990 ] Spikovska and Reid, 1990 ] Atlas et al. 1991 ] and [ Gorman and Sejnowski, ....

.... , Ripley, 1992 ] Mooney et al. 1989, Shavlik et al. 1991 ] Fisher and McKusick, 1989 ] Thrun et al. 1991 ] Kirkwood et al. 1989 ] Tsaptsinos et al. 1990 ] Spikovska and Reid, 1990 ] Atlas et al. 1991 ] and [ Gorman and Sejnowski, 1988 ] The comparisons by [ Weiss and Kapouleas, 1989, Weiss and Kulikowski, 1991 ] found that symbolic algorithms (PVM [ Weiss et al. 1990 ] and CART [ Breiman et al. 1984 ] were more accurate (on most datasets) than back propagation or a number of statistical algorithms (e.g. linear and logistic discriminants) In similar trials involving ID3 ....

[Article contains additional citation context not shown here]

S. M. Weiss and I. Kapouleas. 1989. An empirical comparison of pattern recognition, neural nets and machine learning classification methods (vol 1). IJCAI 89. 20-25 August 1989, Detroit, MI., pp 781 -- 787, San Mateo, CA. Morgan Kaufmann.

....of structural complexity. This paper studies how learning accuracy varies with respect to complexity measures of structured concepts, with special emphasis on concept variation (Rendell Seshu, 1990, 1994) Algorithm evaluations based on collections of realworld datasets seem very pragmatic (Weiss Ioannis, 1989; Shavlik, Mooney, Towell, 1991) However, what is really needed in practice is a characterization of what kind of learner is likely best under the given conditions. Furthermore, datasets often used in empirical comparison are fundamentally similar despite superficial differences (Holte, 1993) ....

Weiss, S. M., & Ioannis, K. (1989). An empirical comparison of pattern recognition, neural nets and machine learning classification methods. In Proc. of the 11th Int. Joint Conf. on Artificial Intelligence, pp. 781--787. Morgan Kaufmann Pub., Inc.

.... from neural networks Running Head: Extracting Rules Submitted (8 91) to Machine Learning Towell Shavlik Extracting Rules 2 1 Introduction Artificial neural networks (ANNs) have proven to be a powerful and general technique for machine learning (Fisher McKusick, 1989; Shavlik et al. 1991; Weiss Kapouleas, 1989; Ng Lippmann, 1990) However, ANNs have several well known shortcomings; perhaps the most significant is that a trained ANN is essentially a black box. That is, determining exactly why an ANN makes a particular decision is all but impossible. Yet, without such an ability, neural networks ....

Weiss, S. M. & Kapouleas, I. (1989). An empirical comparison of pattern recognition, neural nets, and machine learning classification methods. In Proceedings of the Eleventh International Joint Conference on Artificial Intelligence, (pp. 688--693), Detroit, MI.

....suitable for learning have different degrees of structural complexity and hence, different degrees of learning difficulty. Empirical studies of learning accuracy often find no significant difference among systems as a result of using data customized for learning (Holte, 1993; Shavlik et al. 1991; Weiss and Ioannis, 1989). These data are described using attributes carefully selected and or constructed by domain experts. Thus, the given attributes are converted into an important source of knowledge, which simplifies learning greatly and blurs system differences. However, in poorly understood domains, lack of domain ....

Weiss, S. M. and Ioannis, K. (1989). An empirical comparison of pattern recognition, neural nets and machine learning classification methods. In Proc. of the 11th International Joint Conference on Artificial Intelligence, pages 781--787. Morgan Kaufmann Pub., Inc.

....learning, has produced hundreds of computer learning systems. The book further observes that computers are slow and that faster parallel machines are hard to program. Recent research, however, suggests that nonneural inspired machine learning methods are much faster that neural inspired methods (Weiss and Kapouleas, 1989). In addition, many nonnerual methods parallelize easily (Omohundro, 1987) In conclusion, the book tries to draw a distinction between genuine synthetic intelligence and mere artificial intelligence. Its narrow observations do a disservice to the field of artificial intelligence by ignoring ....

Weiss, S. & Kapouleas, I. (1989) An empirical comparison of pattern recognition, neural nets, and machine learning classification methods. In Proceedings of the International Joint Conference on Artificial Intelligence. Detroit, pp.

....have emerged. In the best case such multi method analyses can lead to new insights of the data and to a better understanding of what particular methods are suitable for the problem in question. Different induction methods have frequently been analysed in various domains, e.g. Michie et al. 1994) Weiss and Kapouleas (1989). The results from different comparisons have varied. However, in most cases there does not seem to be a constant overall best method, but different domains have their own characteristics which are met by different methods. In this 1 Copyright 1998, American Association for Artificial Intelligence ....

Weiss, S., and Kapouleas, I. 1989. An Empirical Comparison of Pattern Recognition, Neural Networks, and Machine Learning Classification Methods. In Proceedings of the 11th International Joint Conference on Artificial Intelligence, Detroit, MI, 781-787.

....time and space needed for creating a classifier and finding the classification of a case, respectively. 2.2. 8 Comparison of Different Classification Methods Depending on the domain, different methods perform differently well (for a comparison of different Machine Learning methods see e.g. Weiss and Kapouleas 1989 and Michie, Spiegelhalter, and Taylor 1994) However, the overall performance over all possible domains is about the same for all methods. Schaffer (1994) formulates this more strictly: the overall generalization accuracy of any classifier over all domains and all possible training set test set ....

Weiss, S. M. & I. Kapouleas (1989). An Empirical Comparison of Pattern Recognition, Neural Nets, and Machine Learning Classification Methods. In Proceedings of the Eleventh International Joint Conference on Artificial Intelligence (IJCAI-89), Los Altos, CA, pp. 781--787. Morgan Kaufmann.

....shown in Table 1. Also shown in the table is the accuracy obtained when, for each trainingand test set pair, we take the majority vote of 11 trees when classifying the test set. Note that the accuracy when using the majority voting scheme is consistently higher than when using single SADTtrees. Weiss and Kapouleas (1989) obtained accuracies on this data of 96.7 , 96.0 , and 95.3 with backpropagation, nearest neighbor, and CART, respectively. Their results were generated with leave one out trials, i.e. 150fold cross validation. 92 93 94 95 96 97 98 0 5 10 15 20 25 30 35 Classification Accuracy Number of Trees ....

Weiss, Sholom, and I. Kapouleas, 1989. An empirical comparison of pattern recognition, neural nets, and machine learning classification methods.

....evenly among the concept classes, the set of examples generated was randomly re sampled to obtain a set that was balanced with respect to the frequency of objects from both classes. 96,472 examples were generated for the resulting training set, and 54,841 for the test set. Comparative studies [9, 12] have shown that, during training, decision tree algorithms run significantly faster than connectionist methods. Since one of our objectives was to reduce the time complexity for tracking, we chose C4.5 (version 6) 10] as our learning method. We ran three random trials of C4.5 with windowing ....

Weiss, S., and Kapouleas, I. (1989) An Empirical Comparison of Pattern Recognition, Neural Nets, and Machine Learning Classification Methods. Proceedings of the 11th IJCAI.

....neural networks (ANNs) have proven to be a powerful and general technique for machine learning. For example, a host of empirical comparisons indicate that ANNs are at least as effective at generalizing from training to testing examples as any of several common symbolic machine learning algorithms [Atlas89, Fisher89, Shavlik91, Weiss89]. This generalization aptitude results from the ability of ANNs to train hidden units to form useful intermediate representations. In other words, the success of ANNs results from their ability to use hidden units as loci for constructive induction. However, ANNs tend to be black boxes after ....

Weiss, S. M. and Kapouleas, I. (1989). An empirical comparison of pattern recognition, neural nets, and machine learning classification methods. In Proceedingsof the Eleventh International Joint Conferenceon Artificial Intelligence, pages 688--693, Detroit, MI.

....more time for training than other machine learning methods. Training is normally performed by repeatedly presenting the network with instances from a train4 ing set, and allowing it gradually to converge on the best set of weights for the task, using (for example) the back propagation algorithm. Weiss and Kapouleas (1989), Mooney et al. 1989) and Shavlik et al. 1989) report that back propagation s training time is many orders of magnitude greater than training time for algorithms such as ID3, frequently by factors of 100 or more. In addition, neural net al..gorithms have a number of parameters (e.g. the ....

Weiss, S. and Kapouleas, I. (1989) An empirical comparison of pattern recognition, neural nets, and machine learning classification methods.

....accuracy is output by the Composite Fitness Feature Selection algorithm. 133 This algorithm takes a lesson from the established importance of feature selection to accurate statistical pattern recognition and theoretical and applied inductive learning (e.g. Littlestone, 1987; Krzanowski, 1988; Weiss and Kapouleas, 1989; Tan and Schlimmer, 1990; Callan, 1993; Ripley, 1996] Here, however, we use feature selection not as a technique to increase accuracy, but to increase diversity. This algorithm also draws from the NetTalk example of Wolpert [1992] which selected inputs to component classifiers byhand (Section ....

Weiss, S. M. and Kapouleas, I. 1989. An Empirical Comparison of Pattern Recognition, Neural Nets, and Machine Learning Classification Methods. In 11th International Joint Conference on Artificial Intelligence, Detroit, MI. International Joint Conferences on Artificial Intelligence. 781--787.

....used. Contrasting the performance of symbolic and connectionist approaches is not a new theme and has been covered extensively. Atlas et al. 1990) Dietterich, Hild, and Bakiri (1990) Fisher and McKusick (1989) Fisher et al. 1989) Mooney et al. 1989) Shavlik, Mooney, and Towell (1991) and Weiss and Kapouleas (1989) are seven examples of early empirical studies that have set the tone of the general understanding of the approaches relative strengths and weaknesses. The consensus of these seven studies is that the two approaches are more or less equivalent on prediction accuracy, but when noise is added the ....

Weiss, S. and Kapouleas, I. (1989). An empirical comparison of pattern recognition, neural nets, and machine learning classification methods. In Proc. Eleventh International Joint Conference on Artificial Intelligence (pp. 781--787). San Mateo, CA: Morgan Kaufmann.

....be applied to a wide variety of real world functions. Two promising general techniques for inductive learning are Neural Nets and Function Decomposition. The former is a well established field, with many results related to inductive learning (see [Dietterich et al. 1990] Mooney et al. 1989] [Weiss and Kapouleas, 1989], Fisher and McKusick, 1989] etc. The latter is a relative newcomer, and its merits as an inductive learning algorithm are still being evaluated. Knowing which approach is more general may lead to a deeper understanding of what it means for a function from the real world to have a pattern. It ....

Weiss, S. and Kapouleas, I. (1989). An empirical comparison of pattern recognition, neural nets, and machine learning classification methods. In Sridharan, N., editor, IJCAI-89 Proceedings of the Eleventh International Joint Conference on Artificial Intelligence., volume 1, pages 781--7.

.... which algorithms are better (and under what conditions) Inductive concept learning research is actively proceeding through this phase, as demonstrated in the papers by Fisher (1987) Schlimmer Fisher (1986) Utgoff (1988) Quinlan (1988) Mingers (1989) Mooney, Shavlik, Towell, Gove (1989) Weiss Kapouleas (1989), Fisher McKusick (1989) A fifth phase of research, which may proceed independently of phases three and four, is the theoretical analysis of the task and the methods. For inductive concept learning, there has been an explosion of activity in the past five years, as represented by Valiant ....

Weiss, S., & Kapouless, I. (1989). An empirical comparison of pattern recognition, neural nets, and machine learning classification methods. IJCAI-89: Eleventh International Joint Conference on Artificial Intelligence (pp. 781--787). Detroit, MI: Morgan Kaufmann.

....Irises. This is Fisher s famous iris data, which has been extensively studied in the statistics and machine learning literature. The data consists of 150 examples, where each example is described by four numeric attributes. There are 50 examples of each of three different types of iris flower. Weiss and Kapouleas (1989) obtained accuracies of 96.7 and 96.0 on this data with back propagation and 1 NN, respectively. Housing Costs in Boston. This data set, also available as a part of the UCI ML repository, describes housing values in the suburbs of Boston as a function of 12 continuous attributes and 1 binary ....

Weiss, S., & Kapouleas, I. (1989). An empirical comparison of pattern recognition, neural nets, and machine learning classification methods. In Proceedings of the 11th International Joint Conference of Artificial Intelligence, pp. 781--787. Detroit, MI. Morgan Kaufmann.

....a typical plot of the relative error vs. the number of pseudo classes (Weiss Indurkhya, 1993b) As the number of partitions increases, results improve until they reach a relative plateau and deteriorate somewhat. Similar complexity plots can be found for other models, for example neural nets (Weiss Kapouleas, 1989). The MARS procedure has several adjustable parameters. 6 For the parameter mi, values tried were 1 (additive modeling) 2, 3, 4 and number of inputs. For df, the default value of 3.0 was tried as well the optimal value estimated by cross validation. The parameter nk was varied from 20 to 100 in ....

Weiss, S., & Kapouleas, I. (1989). An empirical comparison of pattern recognition, neural nets, and machine learning classification methods. In International Joint Conference on Artificial Intelligence, pp. 781--787 Detroit, Michigan.

No context found.

S.M. WEISS and I. KAPOULEAS, 1989. An Empirical Comparison of Pattern Recognition, Neural Nets, and Machine Learning Classification Methods, in Proceedings of the Eleventh International Joint Conference on Artificial Intelligence, Morgan Kaufmann, San Mateo, CA, 688 -- 693.

No context found.

REFERENCES 283 Weiss, S. M. and Kapouleas, I. (1989). An empirical comparison of pattern recognition, neural nets and machine learning classification methods (vol 1). In IJCAI 89: proceedings of the eleventh international joint conference on artificial intelligence, Detroit, MI, pages 781--787, San Mateo, CA. Morgan Kaufmann.

No context found.

WEISS,S.M. & KAPOULEAS,I.(1989). An Empirical Comparison of Pattern Recognition, Neural Nets, and Machine Learning Classification Methods. Proceedings of the 11th International Joint Conference on Artificial Intelligence, Instance-Based Prototypical Learning of Set Valued Attributes 23 IJCAI-89,pp. 781--787.

No context found.

Weiss, S., and Kapouleas, I. (1989) An Empirical Comparison of Pattern Recognition, Neural Nets, and Machine Learning Classification Methods. Proceedings of the 11th IJCAI.

No context found.

Weiss, S. M. and Kapouleas, I. (1989). An empirical comparison of pattern recognition, neural nets, and machine learning classification methods. In Proceedings of the Eleventh International Joint Conference on Artificial Intelligence, pages 688--693, Detroit, MI.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC