Hecht-Nielsen, R. Applications of Counterpropagation Networks, Neural Networks, Vol. 1, 1988, pp. 131-139.  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... [6] self organizing map (SOM) 24] or Hopfield [20] networks; do not require that the output value for a training sample be provided at the time of training, while supervised learning systems; such as backpropagation (multi layer perceptron) radial basis function (RBF) 35] counterpropagation [18], or fuzzy ARTMAP [7] a supervised learning extension of the ART Heuristics Principles for the Design of Artificial Neural Networks Page 15 learning method) networks; require that a known output value for all training samples be provided to the ANN. Unsupervised learning methods determine ....

Hecht-Nielsen, R. Applications of Counterpropagation Networks, Neural Networks, Vol. 1, 1988, pp. 131-139.

....very similar to the NNC, and can be used as a classifier. As illustrated in Fig. 1, Kohonen (1990, 1988) and Kohonen et.al. 1988) have mapped the NNC to a neural network, which they have named learning vector quantization (LVQ) They have also suggested a learning rule for training the network. Hecht Nielsen (1987, 1988) developed the two layer counter propagation network (CPN) that combines the Kohonen self organization and Grossberg outstar algorithms. The LVQ and CPN networks are then isomorphic to types of NNC. In this paper we develop techniques for optimizing the NNC through the use of a sigma pi ....

Hecht-Nielsen, R. (1988). Applications of counterpropagation network. Neural Networks, Vol.1, pp.131-139.

....and learning algorithms (supervised bp and unsupervised self organisation in soms) Below, we compare bp som with three other types of hybrid architectures and learning algorithms, viz. counterpropagation networks, radial basis function (rbf) networks, and artmap. In a counterpropagation network (Hecht Nielsen, 1988), the first part of the network implements unsupervised competitive learning by dividing the input space into a Voronoi tessellation, with intermediary units representing the centres of the Voronoi cells. After competitive learning, the activations of the output units are trained in the second ....

Hecht-Nielsen, R. (1988). Applications of counterpropagation networks. Neural Networks, 1, 131--139.

....algorithm used for learning the Classification System and a penalty function to preserve the verification of a feasibility property in the feature subset being evaluated. Brill et al. 9] propose a GA with punctuated equilibria (GAPE) to select variables for a neural network classifier [33]. GAPE is based on the following basic ideas: It uses, like Siedlecki and Sklansky in [71] a binary coding with fixed length to indicate the present variables with an 1, and those not present with a 0. Each individual receives a scoring, which is a linear combination of the error and the ....

....ics.uci.edu in the directory pub machinelearning databases. Gonzlez and Prez [30] show the test results obtained by SLAVE on the IRIS example base, using as error estimation technique random resampling, comparing them with those obtained with other classical learning algorithms: C4.5 [69] and BP [33]. Table 2. Test results for IRIS. Algorithm Test SLAVE 95.43 C4.5 91.13 BP 91.56 One of the more important characteristics of this learning algorithm is the high linguistic description level of the obtained RB. The coding scheme used, along with the genetic operators, make possible to obtain ....

Hecht-Nielsen, R. (1989), "Applications of counterpropagation networks," Neural Networks, Vol. 1, pp. 131-139.

....In our work we used a regression method and principle component analysis (PCA) to reduce the parameters. After this reduction we compared the results of different neural network algorithms to achieve best results. In detail, we tried Radial Basis Functions (RBF) Pog89] Counterpropagation [HN88] and Feedforward networks like Backpropagation (BackProp) Rum86] and Resilient Propagation (RProp) Bra93] For all networks, we first trained with half of the data samples (2430) and all components (96) and used the rest of the 4860 data samples for the validation set. After that we reduced the ....

Robert Hecht-Nielsen. Applications of counterpropagation networks. Neural Networks, 1(2):131--139, 1988.

....example in Section 8. The originally proposed counterpropagation network [6] architecture involved a two way flow (counter flow) of information intended to create a two directional lookup table. A simplified unidirectional network was introduced at the same time, and expanded upon thereafter[7], and is the one used herein. The simplified network is composed of two layers. The first layer comprises competitive instars trained using Kohonen learning. Outstars are emitted from the first layer to form the second layer, and the outstars are trained using Grossberg s outstar learning ....

Hecht-Nielsen, R., "Applications of counterpropagation networks", Neural Networks, 1, 131-139, 1988.

....such as K means [91] and show that neural network methods are able to outperform traditional techniques in this application area in terms of both solution quality and speed. Self organizing neural network architectures such as Kohonen s Feature Map [103] and the Counter propagation network [78] are also suitable for clustering. 30] 3.4 Cutting Stock and Packing Problems This category of problems includes sub classes such as binpacking, knapsack, and cutting problems. Solutions to these problems are important because many of these problems find application in industry. Takada et al. ....

R. HECHT-NIELSEN, 1988. Applications of Counterpropagation Networks, Neural Networks 1, 131--139.

....the image. Noise Both black and white noise can be added to the images with a probability between 0.05 and 0.4. Black and white noise can also be mixed in the same data set. Table 3: Value range of IMAGE5 parameters Counterpropagation model Implementation of the CP algorithm is done according to [3, 2]. The CP algorithm is a three layers feed forward version based on a Kohonen linear associator [5, 6] and Grossberg outstar neurons. The implemented CP network runs in accretive mode. 2.5 Specification of tests From the general hypothesis mentioned in the introduction, several specified ....

Robert Hecht-Nielsen. Application of counterpropagation networks. Neural Networks, 1(2):131--139, 1988.

....in a local optimum earlier or suffer from overgeneralization. In this case our algorithm would even find a better solution. See our experiments for what we found in practice. 6. 3 Related work An early algorithm that combines a SOM with a feedforward network was the Counterpropagation network [Hecht Nielsen, 1988, Hertz et al. 1991] The first two layers are used to code an input manifold and a output manifold. Associative connections between the two layers are learned with the delta rule; the layers are fully connected with each other. Algorithms in which the second layer is connected partially to parts ....

Hecht-Nielsen, R. (1988). Applications of the counterpropagation network. Neural Networks, 1:131--139.

....] use of different objective functions, Baxt s [ 1992 ] method of training networks on different tasks, and Perrone s [ 1992 ] tree structured neural networks. Our method of combining competitive (unsupervised) and backpropagation (supervised) learning is similar to a number of other approaches [ Hecht Nielsen, 1988; Huang and Lippmann, 1988; Moody and Darken, 1989 ] The main difference between our work and this previous research is that we use a hybrid unsupervised supervised network as a method for producing networks that are very effective when used in combination, while the others focused on producing ....

....region of input space; the main difference is that in our work we use competitive learning and our examples to select these regions rather than trying to select them randomly. Our work also relates to hybrid systems that mix levels of unsupervised and supervised learning in neural networks [ Hecht Nielsen, 1988; Huang and Lippmann, 1988 ] One difference in our work is that we perform our unsupervised learning among the categories separately. We also transform the results from competitive learning using our weight multiplier producing large initial weights. Finally, we install the results of ....

R. Hecht-Nielsen. Applications of counterpropagation networks. Neural Networks, 1:131--139, 1988.

....for parallel processing becomes available, the first problem will be solved. Several neural networks which are isomorphic to NNC s have been developed to attack the second problem. These include the learning vector quantization (LVQ) of Kohonen [3] the counter propagation network of HechtNielsen [4], the adaptive clustering network of Barnard and Casasent [5] and the nearest neighbor isomorphic network (NNIN) of Yau and Manry [6] In this paper we discuss properties of product units that allow them to substitute for Min units, in the NNIN. We compare its performance to that of LVQ2.1 for ....

R. Hecht-Nielsen, "Applications of counterpropagation network", Neural Networks, Vol.1, 1988, pp.131-139.

....to this complex mapping problem. The ANN used for this research is based on the Multi Layer Perceptron (MLP) using the popular back propagation training algorithm as discussed in [13] Other ANN approaches can also be employed, for example the counterpropagation network described by Hecht Nielsen [6]. 5.1 The Back Propagation Algorithm Back propagation applies a gradient descent technique which attempts to minimize the error between desired and actual network output based on a set of inputoutput pairs known as the training set. The network consists of a layer of input nodes, multiple ....

R. Hecht-Nielsen, "Applications of Counterpropagation Networks," Neural Networks, vol. 1, no. 2, pp. 131--139, 1988.

....Vector quantization is widely used as a general technique for data compression. Areas of application include image coding [1, 2] speech coding [3, 4] and speech recognition [5, 6] Some approaches that are based on neural network models have also been proposed to quantize data adaptively [7, 8, 9, 10]. The basic idea of vector quantization is to represent a data manifold M by a finite set of n codebook vectors w i 2 K ; i = 1; 2; n, in a K dimensional space. A data vector v 2 M is said to be represented (approximately) by a codebook vector w i if the Euclidean norm kv Gamma w i k is ....

R. Hecht-Nielsen (1988). Applications of counterpropagation networks. Neural Networks, 1:131-- 141.

No context found.

R. Hecht-Nielsen, "Applications of Counterpropagation Networks," Neural Networks, 1(2), 131--139 (1988).

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