K. Ng and R.P. Lippmann. Practical characteristics of neural network and conventional pattern classiers. In J.E. Moody R.P. Lippmann and D.S. Touretzky, editors, Advances in Neural Information Processing Systems-3, pages 970-976, 1991.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....through examples Thus, they do not require a good apriori mathematical model for the underlying physical characteristics. A good review of probabilistic, hyperplane, kernel and exemplar based classi ers that discusses the relative merit of various schemes within each category, is available in [25]. It is observed that, if trained and used properly, several neural networks show comparable performance over a wide variety of classi cation problems, while providing a range of trade o s in training time, coding complexity and memory requirements. Some of these networks, including the ....

K. Ng and R.P. Lippmann. Practical characteristics of neural network and conventional pattern classiers. In J.E. Moody R.P. Lippmann and D.S. Touretzky, editors, Advances in Neural Information Processing Systems-3, pages 970-976, 1991.

....such as fuzzy logic can be incorporated into a neural network classifier for applications with little training data. A good review of probabilistic, hyperplane, kernel and exemplar based classifiers that discusses the relative merit of various schemes within each category, is available in [31]. Although neural networks do not require geometric models, they do do require that the set of examples used for training should come from the same (possibly unknown) distribution as the set used for testing the networks, in order to provide valid generalization and good performance on ....

K. Ng and R. Lippmann, "Practical characteristics of neural network and conventional pattern classifiers," in Neural Information Processing Systems (J. M. R.P. Lippmann and D. Touretzky, eds.), pp. 970--976, 1991.

....networks for speech recognition. Early work used recurrent neural networks for representation of temporal context but after the introduction of time delay neural networks by Waibel at al. 51] feedforward networks are also used for phoneme recognition. Lee and Lippmann [25] and Ng and Lippmann [35] for the same two artificial and two speech tasks compare a large number of conventional and neural pattern classifiers. Comparison of distance based classifiers, single and multilayer perceptrons and radial basis function networks for phoneme recognition is given in [19] The recent book by ....

Ng, K, Lippmann, R. (1991) "Practical Characteristics of Neural Network and Conventional Pattern Classifiers," in R. Lippmann, J. Moody, D. Touretzky (Eds.), Advances in Neural Information Processing Systems 3, 970--976, Morgan Kaufmann.

....two reviews on using neural networks for speech recognition. Early work used recurrent neural networks for representation of temporal context but after the introduction of time delay neural networks by Waibel at al. 51] feedforward networks are also used for phoneme recognition. Lee and Lippmann [25], and Ng and Lippmann [35] for the same two artificial and two speech tasks compare a large number of conventional and neural pattern classifiers. Comparison of distance based classifiers, single and multilayer perceptrons and radial basis function networks for phoneme recognition is given in ....

Lee Y., Lippmann, R. (1990) "Practical Characteristics of Neural Network and Conventional Pattern Classifiers on Artificial and Speech Problems," in D. Touretzky (Ed.) Advances in Neural Information Processing Systems 2, 168--177, Morgan Kaufmann.

....obtained by individual classifiers on this data set are compared to those obtained using two proposed evidence combination techniques in Section 3. Some related work is elaborated upon in [3, 2. STATIC ANN CLASSIFIERS Our experiences, corroborated by those of several other researchers (see [6] for example) show that classification error rates are similar across different ANN classifiers when they are powerful enough to form minimum error decision regions, if they are properly tuned, and when sufficient training data is available. Practical characteristics such as training time, ....

K. Ng and R.P. Lippmann. Practical characteristics of neural network and conventional pattern classifiers. In Advances in Neural Information Processing Systems -3, pages 970--976, 1991.

.... the design of a detection and classification system that uses a hybrid of ANN and statistical pattern recognition techniques tailored to recognizing short duration oceanic signals [10] 2 Overview of ANN Classifiers Used Our experiences, corroborated by those of several other researchers (see [19] for example) show that classification error rates are similar across different ANN classifiers when they are powerful enough to form minimum error decision regions, if they are properly tuned, and when sufficient training data is available. Practical characteristics such as training time, ....

.... their robustness against noise, effects of small training sets, and in their ability to handle high dimensional inputs [2] A good review of probabilistic, hyperplane, kernel and exemplar based classifiers that discusses the relative merit of various schemes within each category, is available in [16, 19]. Comparisons between these classifiers and conventional techniques such as decision trees, K nearest neighbor, Gaussian mixtures, and CART can be found in [19, 23] For this study, we chose the MLP augmented with weight decay strategies [6] and the EBF network, since these classifiers are ....

[Article contains additional citation context not shown here]

K. Ng and R.P. Lippmann. Practical characteristics of neural network and conventional pattern classifiers. In Advances in Neural Information Processing Systems -3, pages 970--976, 1991.

....then discarded. Accordingly, I call the new algorithm the Incremental Delta Bar Delta (IDBD) algorithm. The IDBD algorithm can be used to accelerate learning even on single problems, and that is the primary way in which its predecessors have been justified (e.g. Jacobs 1988; Silva Almeida 1990; Lee Lippmann 1990; Sutton 1986) but its greatest significance I believe is for nonstationary tasks or for sequences of related tasks, and it is on the former that I test it here. The IDBD Algorithm The IDBD algorithm is a meta learning algorithm in the sense that it learns the learning rate parameters of an ....

Lee, Y. & Lippmann, R.P. (1990) Practical characteristics of neural network and conventional pattern classifiers on artificial and speech problems. In Advances in Neural Information Processing Systems 2, D.S. Touretzky, Ed., 168-- 177.

.... relates these properties to Bayesian decision making and to information theoretic results, has emerged recently [19, 20] A good review of probabilistic, hyperplane, kernel and exemplar based classifiers that discusses the relative merit of various schemes within each category, is available in [13, 17, 23]. Comparisons between these classifiers and conventional techniques such as decision trees, K nearest neighbor, Gaussian mixtures, and CART can be found in [23, 25] It is seen that most of these networks show comparable performance over a wide variety of classification problems, while providing a ....

.... kernel and exemplar based classifiers that discusses the relative merit of various schemes within each category, is available in [13, 17, 23] Comparisons between these classifiers and conventional techniques such as decision trees, K nearest neighbor, Gaussian mixtures, and CART can be found in [23, 25]. It is seen that most of these networks show comparable performance over a wide variety of classification problems, while providing a range of trade offs in training time, coding complexity and memory requirements [9, 23] Neural networks are not magical . They do require that the set of ....

[Article contains additional citation context not shown here]

K. Ng and R.P. Lippmann. Practical characteristics of neural network and conventional pattern classifiers. In Advances in Neural Information Processing Systems -3, pages 970--976, 1991.

....for speech recognition. Early work used recurrent neural networks for representation of temporal context but after the introduction of Time Delay Neural Networks by Waibel at al. 37] a feed forward network can also be used for phoneme recognition. Lee and Lippmann [23] and Ng and Lippmann [30] for the same two artificial and two speech tasks compare a large number of conventional and neural pattern classifiers. B. Classification Methods The classifiers considered are: 1. Kernel estimators [36] like k nearest neighbor (k nn) 11] Parzen windows, generalized k nn, and Grow and Learn ....

Ng K, Lippmann R. Practical Characteristics of Neural Network and Conventional Pattern Classifiers. In R. Lippmann, J. Moody, D. Touretzky (Eds.), Advances in Neural Information Processing Systems 3 (pp. 970--976), Morgan Kaufmann, 1991.

....on using neural networks for speech recognition. Early work used recurrent neural networks for representation of temporal context but after the introduction of Time Delay Neural Networks by Waibel at al. 37] a feed forward network can also be used for phoneme recognition. Lee and Lippmann [23], and Ng and Lippmann [30] for the same two artificial and two speech tasks compare a large number of conventional and neural pattern classifiers. B. Classification Methods The classifiers considered are: 1. Kernel estimators [36] like k nearest neighbor (k nn) 11] Parzen windows, ....

Lee Y, Lippmann R. Practical Characteristics of Neural Network and Conventional Pattern Classifiers on Artificial and Speech Problems. In D. Touretzky (Ed.) Advances in Neural Information Processing Systems 2 (pp. 168--177), Morgan Kaufmann, 1990.

....characterization of short duration acoustic signals is not available yet. There are several neural networks that show comparable performance over a wide variety of classification problems, while providing a range of trade offs in training time, coding complexity and memory requirements [7, 8]. Some of these networks, including the multilayered perceptron when augmented with weight decay strategies [9] and the elliptical basis function network introduced in this paper, are quite insensitive to noise and to irrelevant inputs [10] Moreover, a firmer theoretical understanding of the ....

....memory intensive. A novel hybrid network that achieves these objectives is also discussed in Section 3.3, and forms our second classifier candidate. Higher order networks based on the GMDH algorithm often require long training times as well as large amount of memory to yield comparable error rates [8]. Polynomial networks based on Volterra series expansion [48, 49] show fairly stable, single layer learning, but the number of weights involved grows exponentially with the order of the network. We have recently proposed a higher order network called the Pi Sigma network, that is able to maintain ....

K. Ng and R.P. Lippmann. Practical characteristics of neural network and conventional pattern classifiers. In Advances in Neural Information Processing Systems -3, pages 970--976, 1991.

.... and classification system that uses a hybrid of ANN and statistical pattern recognition techniques tailored to recognizing short duration oceanic signals [Gho91, GBC92, GDB92, BG92] 2 Overview of ANN Classifiers Used Our experiences, corroborated by those of several other researchers (see [NL91] for example) show that classification error rates are similar across different ANN classifiers when they are powerful enough to form minimum error decision regions, if they are properly tuned, and when sufficient training data is available. Practical characteristics such as training time, ....

.... robustness against noise, effects of small training sets, and in their ability to handle high dimensional inputs [BG92] A good review of probabilistic, hyperplane, kernel and exemplar based classifiers that discusses the relative merit of various schemes within each category, is available in [Lip89, NL91]. Comparisons between these classifiers and conventional techniques such as decision trees, K nearest neighbor, Gaussian mixtures, and CART can be found in [NL91, WK91] For this study, we chose the MLP and the EBF network, since these classifiers are comparatively insensitive to noise and to ....

[Article contains additional citation context not shown here]

K. Ng and R.P. Lippmann. Practical characteristics of neural network and conventional pattern classifiers. In Advances in Neural Information Processing Systems -III, pages 970--976, 1991.

....It is natural then to expect cross fertilization, in which ideas developed in one field may also contribute to methods used in another. One topic of recent interest in connectionist learning is the automatic selection of learning rate or gain parameters during learning (Jacobs, 1988; Lippmann Lee, 1990; Sutton, 1992; Gluck, Glauthier Sutton, 1992; cf. Kesten, 1958) Such dynamic learning rate (DLR) methods have been shown to speed convergence on a variety of learning tasks, particularly those with many irrelevant input features and in which the task is non stationary (the correct solution ....

Lee, Y. & Lippmann, R.P. (1990) Practical characteristics of neural network and conventional pattern classifiers on artificial and speech problems. In Advances in Neural Information Processing Systems 2, D.S. Touretzky, Ed., 168--177, Morgan Kaufmann.

No context found.

Y. Lee and R.P. Lippmann. Practical characteristics of neural network and conventional pattern classifiers on artificial and speech problems. In Neural Information Processing Systems 2, pages 168--177, 1990.

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