A New Adaptive Merging and Growing Algorithm for Designing Artificial Neural Networks

This paper presents a new algorithm, called adaptive merging and growing algorithm (AMGA), in designing artificial neural networks (ANNs). This algorithm merges and adds hidden neurons during the training process of ANNs. The merge operation introduced in AMGA is a kind of a mixed mode operation, which is equivalent to pruning two neurons and adding one neuron. Unlike most previous studies, AMGA puts emphasis on autonomous functioning in the design process of ANNs. This is the main reason why AMGA uses an adaptive not a predefined fixed strategy in designing ANNs. The adaptive strategy merges or adds hidden neurons based on the learning ability of hidden neurons or the training progress of ANNs. In order to reduce the amount of retraining after modifying ANN architectures, AMGA prunes hidden neurons by merging correlated hidden neurons and adds hidden neurons by splitting existing hidden neurons. The proposed AMGA has been tested on a number of benchmark problems in machine learning and ANNs, including breast cancer, Australian credit card assessment, and diabetes, gene, glass, heart, iris, and thyroid problems. The experimental results show that AMGA can design compact ANN architectures with good generalization ability compared to other algorithms.     

Lymphoid/nonlymphoid compartmentalization of donor leukocyte chimerism in rat recipients of heart allografts, with or without adjunct bone marrow

A spectrum of oxidative lesions was observed in a bacteriophage-based model system that is very sensitive to the photodynamic activity of selected dyes. When suspensions of the intact bacteriophage Q beta were exposed to methylene blue plus light (MB + L), inactivating events, or "hits" occurred that were oxygen-dependent and that were associated with the formation of several specific lesions: (1) carbonyl moieties on proteins, (2) 8-oxo-7,8-dihydroguanine (8-oxoGua), and (3) single-strand breaks (ssb) in the RNA genome and (4) RNA-protein crosslinks. Formation of carbonyl groups associated with protein in the Q beta phage preparation correlated positively with photoinactivation of the phage with increasing doses of either of the sensitizers MB or rose bengal. Strand breaks in the Q beta genomic RNA were observable at high MB concentrations but appeared not to be significant at the lower concentrations of MB, as full-length Q beta RNA was observable well beyond the 99% inactivation point in MB dosage. It was shown that the number of 8-oxoGua lesions were unlikely to be sufficient to account for the number of lethal events. Following exposure to MB + L, crosslink formation between Q beta RNA and protein was observed by virtue of the location of RNA at the interface of phenol-aqueous extractions of phage suspensions. A significant increase over background of RNA-protein complexes (including full-length Q beta RNA) was observed at the lowest concentration of MB tested (0.5 microM), which corresponded roughly to an average of 2 lethal hits per phage or approximately 13% survival compared to the zero MB control (100% survival). Due to its close correlation with Q beta inactivation and its expected lethality, RNA-protein crosslink formation may be important as an inactivating lesion in bacteriophage Q beta following MB + L exposure.     

Recognition of camera-captured low-quality characters using motion blur information

Camera-based character recognition has gained attention with the growing use of camera-equipped portable devices. One of the most challenging problems in recognizing characters with hand-held cameras is that captured images undergo motion blur due to the vibration of the hand. Since it is difficult to remove the motion blur from small characters via image restoration, we propose a recognition method without de-blurring. The proposed method includes a generative learning method in the training step to simulate blurred images by controlling blur parameters. The method consists of two steps. The first step recognizes the blurred characters based on the subspace method, and the second one reclassifies structurally similar characters using blur parameters estimated from the camera motion. We have experimentally proved that the effective use of motion blur improves the recognition accuracy of camera-captured characters.     

Diversity Guided Evolutionary Programming: A novel approach for continuous optimization

Avoiding premature convergence to local optima and rapid convergence towards global optima has been the major concern with evolutionary systems research. In order to avoid premature convergence, sufficient amount of genetic diversity within the evolving population is considered necessary. Several studies have focused to devise techniques to control and preserve population diversity throughout the evolution. Since mutation is the major operator in many evolutionary systems, such as evolutionary programming and evolutionary strategies, a significant amount of research has also been done for the elegant control and adaptation of the mutation step size that is proper for traversing across the locally optimum points and reach for the global optima. This paper introduces Diversity Guided Evolutionary Programming, a novel approach to combine the best of both these research directions. This scheme incorporates diversity guided mutation, an innovative mutation scheme that guides the mutation step size using the population diversity information. It also takes some extra diversity preservative measures to maintain adequate amount of population diversity in order to assist the proposed mutation scheme. An extensive simulation has been done on a wide range of benchmark numeric optimization problems and the results have been compared with a number of recent evolutionary systems. Experimental results show that the performance of the proposed system is often better than most other algorithms in comparison on most of the problems.     

A quick search method for audio and video signals based on histogram pruning

This paper proposes a quick method of similarity-based signal searching to detect and locate a specific audio or video signal given as a query in a stored long audio or video signal. With existing techniques, similarity-based searching may become impractical in terms of computing time in the case of searching through long-running (several-days' worth of) signals. The proposed algorithm, which is referred to as time-series active search, offers significantly faster search with sufficient accuracy. The key to the acceleration is an effective pruning algorithm introduced in the histogram matching stage. Through the pruning, the actual number of matching calculations can be reduced by 200 to 500 times compared with exhaustive search while guaranteeing exactly the same search result. Experiments show that the proposed method can correctly detect and locate a 15-s signal in a 48-h recording of TV broadcasts within 1 s, once the feature vectors are calculated and quantized. As extentions of the basic algorithm, efficient AND/OR search methods for searching for multiple query signals and a feature dithering method for coping with signal distortion are also discussed.     

Bubble growth and transient heat transfer at the onset of boiling

An experimental study was conducted to investigate transient local heat transfer around a bubble at onset of boiling on a thin glass heating plate immersed in saturated n-hexane at low pressure. Eight rapid response Cu-Ni thermocouples consisting of a vacuum deposited thin film were used to measure the temperature change of the heating surface. Simultaneous high-speed video photographs were also obtained. The surface temperatures near a nucleation site decreased rapidly owing to the evaporation of a thin layer (microlayer) of liquid formed beneath the bubble in the early period and the rate of bubble growth increased with increasing incipient boiling superheat (ΔT_IB). The thickness of the microlayer decreased markedly with increasing ΔT_IB. © 1998 Scripta Technica, Heat Trans Jpn Res, 26(7): 484–492, 1997     

The Fatty Acid Composition in Tuna (Bonito,Euthynnus pelamis) Caught at Three Different Localities from Tropics to Temperate

The present paper describes the effect of habitat on the fatty acid composition of the lipid of bonito (Euthynnus pelamis), which was caught at three different localities, Philippine Sea (the tropical zone; seawater temperature at the fishing ground was 27·8°C), East China Sea (the subtropical zone; seawater temperature was 29·7°C), and the Pacific coast of Japan (the temperate zone; sea-water temperature was 20·3°C). The total lipids of various organs and stomach contents were extracted and their fatty acid composition analysed by gas chromatography (GC). Docosahexaenoic acid (DHA; 22: 6n-3) was the major unsaturated fatty acid in the lipid of all specimens examined from all localities. The mean DHA content accounted for more than 25% (mean±standard error: 26·0±0·6%) of the total fatty acids (TFA) in the lipids of all organs, a lipid profile markedly different from that of other fish species whose fatty acid composition is generally variable. Particularly, the mean DHA content of lipids in bonito caught in the northern sea-area sample (the temperate zone) was always high (28·4±0·7% TFA) in the lipid of every organ. Because the DHA contents of lipids of the stomach contents fluctuated between 9·9 and 31·9% TFA, bonito did not simply incorporate the fatty acid profile of the lipids of its prey fishes, but selectively accumulated the DHA. Though the mean DHA content in the lipid of all bonito from the tropical to the temperate zone was markedly higher than other marine fish species such as sardines and herrings, there was a small difference between those in the northern (28·4±0·7% TFA) and southern samples (the subtropical and tropical samples, 24·7±0·8% TFA) (P<0·05). It is suggested that the difference between them may be due to environmental effects, for example, the temperature of the seawater and the fatty acid composition of the lipids of prey organisms. © 1997 SCI.     

Fucoxanthin Prevents Pancreatic Tumorigenesis in C57BL/6J Mice That Received Allogenic and Orthotopic Transplants of Cancer Cells

Fucoxanthin (Fx) is a marine carotenoid with anti-inflammatory and anti-cancer properties in various animal models of carcinogenesis. However, there is currently no information on the effects of Fx in animal models of pancreatic cancer. We investigated the chemopreventive effects of Fx in C57BL/6J mice that received allogenic and orthotopic transplantations of cancer cells (KMPC44) derived from a pancreatic cancer murine model (Ptf1a^Cre/+; LSL-kras^G12D/+). Using microarray, immunofluorescence, western blot, and siRNA analyses, alterations in cancer-related genes and protein expression were evaluated in pancreatic tumors of Fx-administered mice. Fx administration prevented the adenocarcinoma (ADC) development of pancreatic and parietal peritoneum tissues in a pancreatic cancer murine model, but not the incidence of ADC. Gene and protein expressions showed that the suppression of chemokine (C-C motif) ligand 21 (CCL21)/chemokine receptor 7 (CCR7) axis, its downstream of Rho A, B- and T-lymphocyte attenuator (BTLA), N-cadherin, αSMA, pFAK(Tyr³⁹⁷), and pPaxillin(Tyr³¹) were significantly suppressed in the pancreatic tumors of mice treated with Fx. In addition, Ccr7 knockdown significantly attenuated the growth of KMPC44 cells. These results suggest that Fx is a promising candidate for pancreatic cancer chemoprevention that mediates the suppression of the CCL21/CCR7 axis, BTLA, tumor microenvironment, epithelial mesenchymal transition, and adhesion.     

Reduction of nitrogen monoxide to nitrogen at gas diffusion electrodes with noble metal catalysts

The electrochemical reduction of NO in alkaline solutions was investigated at gas diffusion electrodes with various metal (Ru, Rh, lr, Pd and Pt) catalysts at various NO flow rates. Reduction currents are observed at potentials more negative than 0.95 V, which increase with the decrease in potential and also with increasing gas flow rate. The faradaic efficiencies of N2O formation decrease with decreasing NO flow rate and with decrease in potential. The faradaic efficiencies of N2 formation increase with decreasing flow rate and with decrease in potential. The reduction of NO to N2 at a flow rate of 5mlmin–1 occurs selectively at potentials more negative than 0.1V; the faradaic efficiency of N2 formation is approximately 95 at Pd catalysts.Electricity production and NO decomposition can be carried out simultaneously using an H2NO fuel cell reactor. The faradaic efficiency of N2 formation at a flow rate of 5mlmin–1 is approximately 80 at a cell voltage of 0.25 V.     

Frequency domain discussions on PD signal propagation characteristics in GIS

The frequency dependency of PD signal propagation in GIS was investigated at an underground substation. Injecting dummy pulses into a GIS through its internal electrode, frequency spectra were measured by a spectrum analyzer at various points of the GIS. It is shown that this measurement is very reproducible and the signal propagation phenomena are independent of its direction. It is also clarified that the PD signal reduction phenomena are closely related with mode transfer from TEM to TE or TM. The major higher mode seems to be TE₁₁. Pulse waveforms were also measured with 65‐MHz bandpass filter. Their magnitudes are compared with the gain of the 65‐MHz component of the frequency spectra and good agreement is obtained. This implies the possibility of PD location by observing a frequency pattern with several bandpass filters. © 2000 Scripta Technica, Electr Eng Jpn 133(3): 9–17, 2000