N. K. Bose and P. Liang, Neural network fundamentals with graphs, algorithms, and applications, McGraw-Hill, New York, NY, 1996.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....insights from connectionism. It is essential that these be integrated [22] The aim of neural symbolic integration is to explore the advantages that each paradigm presents. Within the features of artificial neural networks are massive parallelism, inductive learning and generalization capabilities [7,13]. On the other hand, symbolic systems can explain their inference process, e.g. through automatic theorem proving, and use powerful declarative languages for knowledge representation [17,19] The Connectionist Inductive Learning and Logic Programming (CIL 2 P ) system [5] is a proposal towards ....

N.K. Bose, P. Liang, Neural Networks Fundamentals with Graphs, Algorithms, and Applications, McGrawHill, New York, 1996.

....with them w.r.t. the underlying sampling distribution. This, however, is a necessary condition and not su cient, as there can be local minima of the error function that meet this requirement. The simplest LVQ algorithm is the winner take all competitive learning algorithm described as follows ([11], Ch.9) Let Q i (n) denote the position of the i th centroid at time n. Given a new observation x(n) one flrst locates the centroid that is closest to it (the winner ) say q j (n) and then moves it a little towards x(n) according to the update q j (n 1) 1 c(n) q j (n) c(n)x(n) ....

....n c(n) 2 1: This algorithm can converge to a local minimum of the error function that may be far from the optimal. There are schemes which, while not guarranteeing convergence to the global minimum, do generally improve upon the above algorithm. One such scheme is the Kohonen LVQ algorithm ([11], Ch.9) wherein one updates not only q j (n) but also the centroids that are its neighbours as per some prespecifled neighbourhood scheme. Various weighting schemes have also been proposed to modulate the extent to which difierent centroids should be moved [18] 20] A general formalism for ....

BOSE N.K., LIANG P., Neural Network Fundamentals with Graphs, Algorithms and Applications, McGraw-Hill, Inc., New York, 1996.

....alar3 number of mutually connected simplenonlinear devices, numerC2 str2j04A3 have been pr2 osedfor the re20N0432j of arCCGA42 arCGA42j associativememor neural networks. In par4N2j0qC in 198 , Hopfield intr duced anener4 function todemonstr0G the behavior of fully connectedneurc networO [1] [7]. Some content addressability and robustness to small erll2 in the input [7] ar r2CC05 by these networ strN4 turN wher all weights of the connections of L nonlinear devices called neurd2 ar assigned, based on each designstrgn25 ,for given set ofpatter4 x k # R L (k =1, N) to bememor2jCq ....

....have been pr2 osedfor the re20N0432j of arCCGA42 arCGA42j associativememor neural networks. In par4N2j0qC in 198 , Hopfield intr duced anener4 function todemonstr0G the behavior of fully connectedneurc networO [1] 7] Some content addressability and robustness to small erll2 in the input [7] ar r2CC05 by these networ strN4 turN wher all weights of the connections of L nonlinear devices called neurd2 ar assigned, based on each designstrgn25 ,for given set ofpatter4 x k # R L (k =1, N) to bememor2jCq In there20CqAG stage, the state of the networ continues to be updated ....

N.K. Bose and P. Liang, Neural network fundamentals with graphs, algorithms, and applications, McGraw-Hill, 1996.

....von Neumann machines are very useful for investigating the capabilities of neural network models, a hardware implementation is essential to fully profit from their inherent parallelism, as well as for real time processing in real world problems. Well known ANN, such as multi layer Perceptrons [1], have proved capable of solving various problems, but because of their relatively high complexity, they are not well suited for a hardware implementation. Thus, we chose the FAST (Flexible Adaptable Size Topology) algorithm [2] which is very well suited for such an implementation. This algorithm ....

P. Liang N.K. Bose. Neural Network Fundamentals with graphs, Algorithms, and Applications. Electrical and Computer Engineering Series. McGraw-Hill, Inc., 1996.

....perform subtasks in a serial manner, co operating by passing the output activation of one module to the input layer of another, whilst others execute subtasks in parallel. The notion of modularity has influenced some researchers to combine a range of learning strategies, known as hybrid learning (Bose Liang, 1996:391) for cognitive modelling. A blend of supervised and unsupervised learning paradigms allows categorisation of patterns into pre defined classes to take place within the same model as the clustering of patterns to reflect their probability distribution. An example of a neural architecture ....

Bose, N. K. & Liang, P. (1996) Neural Network Fundamentals with Graphs, Algorithms, and Applications. McGraw-Hill, Inc.

No context found.

N. K. Bose and P. Liang, Neural network fundamentals with graphs, algorithms, and applications, McGraw-Hill, New York, NY, 1996.

No context found.

Bose, N.K., Liang, P.: Neural network fundamentals with graphs, algorithms, and applications. McGraw-Hill, Inc. (1996)

No context found.

N.K. Bose and P. Liang, Neural Network Fundamentals with Graphs, Algorithms, and Applications, McGraw-Hill, New York, 1996. 4

No context found.

Bose, N.; Liang, P. (1996): Neural Network Fundamentals with Graphs, Algorithms, and Applications. New York et al.

No context found.

N. K. Bose and P. Liang. Neural Network Fundamentals with Graphs, Algorithms and Applications. McGraw-Hill, New York, NY, 1996. 39

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