Pattern Classification Based on RBF Networks with Self-Constructing Clustering and Hybrid Learning

Radial basis function (RBF) networks are widely adopted to solve problems in the field of pattern classification. However, in the construction phase of such networks, there are several issues encountered, such as the determination of the number of nodes in the hidden layer, the form and initialization of the basis functions, and the learning of the parameters involved in the networks. In this paper, we present a novel approach for constructing RBF networks for pattern classification problems. An iterative self-constructing clustering algorithm is used to produce a desired number of clusters from the training data. Accordingly, the number of nodes in the hidden layer is determined. Basis functions are then formed, and their centers and deviations are initialized to be the centers and deviations of the corresponding clusters. Then, the parameters of the network are refined with a hybrid learning strategy, involving hyperbolic tangent sigmoid functions, steepest descent backpropagation, and least squares method. As a result, optimized RBF networks are obtained. With this approach, the number of nodes in the hidden layer is determined and basis functions are derived automatically, and higher classification rates can be achieved. Furthermore, the approach is applicable to construct RBF networks for solving both single-label and multi-label pattern classification problems.

The importance of transmethylation reactions to methionine metabolism in sheep: effects of supplementation with creatine and choline

The influence of administering the methylated products choline and creatine on methionine irreversible loss rate (ILR) and recycling from homocysteine has been investigated in sheep fed close to energy and N equilibrium. Two methods to estimate methionine recycling were compared. The first involved & [U-13C& ]methionine infused as part of a labelled amino acid mixture obtained from hydrolysed algal protein. In this approach the isotope dilution of methionine with all five C atoms labelled (m+5) will represent the ILR which does not recycle through homocysteine, while that which includes molecules with C-l–C-4 labelled will allow for loss of the labelled methyl (5)-C atom and replacement by an unlabelled moiety in the remethylation of homocysteine. The second method involved a combined infusion of [l-13C]- & [S-methyl-2H3& rsqb;methionine. These two approaches gave similar data for methionine ILR which does not include label recycled to the amino acid from homocysteine but differed for recycled methionine fluxes. Consequently the two procedures differed in the calculated extent of homocysteine methylation under control conditions (6 v. 28). These extents of remethylation are within the range observed for the fed human subject, despite the fact that fewer dietary methyl groups are available for the ruminant. Using combined data from the infwions, significant depression of methionine recycling occurred in blood (P <0·05), with a similar trendfor plasma (P = 0·077), when choline plus creatine were infused. Wool growth, assessed by intradermal injection of [35S]cysteine, was not altered by supplementation with the methylated products. From changes in the label pattern of free methionine in aortal, hepatic portal and hepatic venous blood during U-13C-labelled algal hydrolysate infusion, the major sites of homocysteine remethylation appear to be the portal-drained viscera and the liver. This was confirmed by analysis of free methionine enrichments in various tissues following dual infusion of & [1J3C& ]- and & [S-methyl-2H3methionine, with the greatest activities occurring in rumen, jejunum and liver. Of the non-splanchnic tissues examined, only kidney exhibited substantial methionine cycling; none was detected in muscle, heart, lung and skin. The implications of methyl group provision under net prduction conditions are discussed.

Robot vision implementation by high-speed image processor TOSPIX: Battery inspection

SUMMARYThis paper presents an intelligent robot vision system using TOSPIX which has been newly developed to realize frequently-used and time-consuming image processing functions at low-cost and high-speed. The vision system has been studied for use in observing surface information about electric parts (dry batteries), inspecting them and then placing good ones into a given box. Three major robot vision functions are implemented here: object recognition, inspection and position determination by binary and gray-scale image processing techniques. While binary image techniques are used in battery terminal inspection and box position determination gray-scale image processing functions are performed in a label pattern check on a battery surface, front or rear surface determination, and surface defect inspection.