Enhancing the generalization ability of neural networks through controlling the hidden layers

In this paper we proposed two new variants of backpropagation algorithm. The common point of these two new algorithms is that the outputs of nodes in the hidden layers are controlled with the aim to solve the moving target problem and the distributed weights problem. One algorithm (AlgoRobust) is not so insensitive to the noises in the data, the second one (AlgoGS) is through using Gauss–Schmidt algorithm to determine in each epoch which weight should be updated, while the other weights are kept unchanged in this epoch. In this way a better generalization can be obtained. Some theoretical explanations are also provided. In addition, simulation comparisons are made between Gaussian regularizer, optimal brain damage (OBD) and the proposed algorithms. Simulation results confirm that the new proposed algorithms perform better than that of Gaussian regularizer, and the first algorithm AlgoRobust performs better than the second algorithm AlgoGS in the noisy data. On the other hand AlgoGS performs better than the AlgoRobust on the data without noise and the final structure obtained by two new algorithms is comparable to that obtained by using OBD.     

Coincidence timing of Schottky CdTe detectors for tomographic imaging

Schottky CdTe detectors are good candidates for large pixel array imagers. For use of such arrays in Positron Emission Tomography (PET) the coincidence timing of the detectors is very important, since it reduces the background from random coincidences. The coincidence timing of a small 2×2×1 mm³ detector has been measured as 11.6 ns, and thus is not much worse than the timing of BGO scintillators, commonly used in PET.     

Preparation and characterization of dicesium tetravalent plutonium hexanitrate

In experiments on U crystallization in a dissolver solution containing Cs, there is concern that Cs and Pu(IV) nitrate complex are deposited on the UNH crystal in the dissolver solution at the time of the U crystallization. The characteristics of generation of Cs and Pu(IV) nitrate complex with dissolver solution of MOX fuel were examined. This complex was obtained as a precipitate by mixing dissolver solution of MOX fuel and CsNO₃ solution, and was identified as dicesium tetravalent plutonium hexanitrate, Cs₂Pu(NO₃)₆ by concentration analysis and XRD. The precipitate has a tendency to be generated at high HNO₃ concentrations. Thermal analysis shows that the precipitate is stable below 245 °C, and a weight loss of 10.29 ± 0.23% is observed between 245 and 297 °C. This result indicates the decomposition of Cs₂Pu(NO₃)₆ to Cs₂PuO₂(NO₃)₄. With these properties, the UNH crystal should melt at the condition between 60 and 100 °C and be separable from the Cs complex by filtration. This suggests a new method of crystal purification allowing higher decontamination of UNH crystal in the U crystallization process.     

Universal learning network and its application to robust control

Universal learning networks (ULNs) and robust control system design are discussed, ULNs provide a generalized framework to model and control complex systems. They consist of a number of interconnected nodes where the nodes may have any continuously differentiable nonlinear functions in them and each pair of nodes can be connected by multiple branches with arbitrary time delays. Therefore, physical systems which can be described by differential or difference equations and also their controllers can be modeled in a unified way. So, ULNs constitute a superset of neural networks or fuzzy neural networks. In order to optimize the systems, a generalized learning algorithm is derived for the ULNs, in which both the first order derivatives (gradients) and the higher order derivatives are incorporated. The derivatives are calculated by using forward or backward propagation schemes. These algorithms for calculating the derivatives are extended versions of back propagation through time (BPTT) and real time recurrent learning (RTRL) by Williams in the sense that generalized nonlinear functions and higher order derivatives are dealt with. As an application of ULNs, the higher order derivative, one of the distinguished features of ULNs, is applied to realizing a robust control system in this paper. In addition, it is shown that the higher order derivatives are effective tools to realize sophisticated control of nonlinear systems. Other features of ULNs such as multiple branches with arbitrary time delays and using a priori information will be discussed in other papers.     

Fast computation of Q‐value‐based dynamic programming on road networks

One of the essential components of vehicle navigation systems is route planning. The single shortest path problem and multiple shortest path problem have been widely studied for route planning. This paper introduces a Q‐value‐based dynamic programming using the division concept for solving both single and multiple shortest path problems on road networks. The proposed algorithm divides the whole network into different divisions, and the updating of Q values in each division is one stage for searching the optimal routes on road networks. The proposed algorithm can greatly save the computational time without any preprocessing on the road networks. The proposed algorithm is also systematically studied in various sizes of road networks. The simulation results show the efficiency and effectiveness of the proposed algorithm on large‐scale road networks. © 2012 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.     

Robust intelligent PCA‐based face recognition framework using GNP‐fuzzy data mining

Traditional principal component analysis (PCA) based face recognition algorithms have a low recognition accuracy due to the influence of noise and illumination changes. This paper proposes a robust, intelligent PCA‐based face recognition framework in the complicated illumination database when using multiple training images per person (MTIP‐CID). There are mainly two improvements in the proposed method. One is that a face‐recognition‐oriented genetic‐based clustering algorithm is introduced to reduce the influence of a large number of classes on the classification accuracy in the MTIP‐CID. The other is that a classifier based on fuzzy class association rules (FCARs) is applied to mine the inherent relationships between eigenfaces and to improve the robustness of PCA‐based face recognition in noisy environments. Experimental results on the extended Yale‐B database demonstrate that the proposed framework performs better and is more robust against noise compared with other traditional face recognition algorithms, i.e. linear discriminant analysis (LDA) and local binary patterns (LBPs). © 2013 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.     

Germline Whole-Gene Deletion of FH Diagnosed from Tumor Profiling

Hereditary leiomyomatosis and renal cell carcinoma (HL (RCC)) entails cutaneous and uterine leiomyomatosis with aggressive type 2 papillary RCC-like histology. HLRCC is caused by pathogenic variants in the FH gene, which encodes fumarate hydratase (FH). Here, we describe an episode of young-onset RCC caused by a genomic FH deletion that was diagnosed via clinical sequencing. A 35-year-old woman was diagnosed with RCC and multiple metastases: histopathological analyses supported a diagnosis of FH-deficient RCC. Although the patient had neither skin tumors nor a family history of HLRCC, an aggressive clinical course at her age and pathological diagnosis of FH-deficient RCC suggested a germline FH variant. After counseling, the patient provided written informed consent for germline genetic testing. She was simultaneously subjected to paired tumor profiling tests targeting the exome to identify a therapeutic target. Although conventional germline sequencing did not detect FH variants, exome sequencing revealed a heterozygous germline FH deletion. As such, paired tumor profiling, not conventional sequencing, was required to identify this genetic deletion. RCC caused by a germline FH deletion has hitherto not been described in Japan, and the FH deletion detected in this patient was presumed to be of maternal European origin. Although the genotype-phenotype correlation in HLRCC-related tumors is unclear, the patient’s family was advised to undergo genetic counseling to consider additional RCC screening.     

Effect of ultrasonic irradiation on the selective polymorph control in sulfamerazine

The selective polymorph control in sulfamerazine (SMZ) using ultrasonic irradiation was investigated. Firstly, the behavior of polymorphs in the absence of ultrasonic irradiation with various degrees of supersaturation was monitored. As a result, only metastable form-I crystallized at the primary nucleation stage and crystal growth of only form-I was recognized. The nucleation rate of stable form-II was very small, and the transformation from form-I to form-II hardly occurred. Therefore, it is easy to obtain form-I without ultrasonic irradiation, as it does not depend on the level of supersaturation. Secondly, the behavior of polymorphs with various degrees of ultrasonic energy and supersaturation was monitored. As a result, only form-I was obtained in the case of small ultrasonic energy. However, in addition to the nucleation of form-I, an unknown form-III and/or the recognized form-II are induced to nucleate under the influence of the ultrasonic energy in the case of large ultrasonic energy. Eventually, all forms changed to form-II within a short period of time. Therefore, polymorph control of SMZ can be achieved easily by supplying an appropriate amount of ultrasonic energy.     

A new lead alloy current-collector manufactured by a powder rolling process and its corrosion behavior under lead-acid battery conditions

A new powder rolling process for manufacturing current-collector sheets for lead-acid batteries has been developed. Gas-atomized lead–tin and lead–tin–calcium alloy powders obtained by a rapid solidification process in air were employed as raw materials for the powder rolling process. The corrosion behavior of powder-rolled lead–tin alloys with various compositions of tin has been investigated. A dipping corrosion test of square plain sheets of the alloys was performed in H₂SO₄ at 75 °C. The test was repeated up to 20 cycles with each cycle consisting of a controlled 10 mA cm⁻² oxidation current for 6 h and a rest under open circuit voltage for 6 h. The extent of corrosion–elongation and the appearance of the corroded surface of the tested specimens were the main observations. The corrosion–elongation of the corroded sheet of a powder-rolled lead alloy containing 1.5 wt% tin with ca. 200 μm initial thickness was less than 5%, whereas that of the corroded sheet of the cast-rolled lead alloy containing 1.5 wt% tin with the same initial thickness was 25–30% under the same corrosion test conditions. The corroded powder-rolled sheet of the 1.5 wt% tin lead alloy has uniform corrosion, but the cast-rolled sheet of lead alloy containing 1.5 wt% tin was much distorted and was perforated by the corrosion. Intergranular corrosion of the powder-rolled lead–tin alloys was much suppressed as compared with that of the cast-rolled lead–tin alloys.     

Inhibition by dexamethasone of antigen-induced c-Jun N-terminal kinase activation in rat basophilic leukemia cells

Antigen stimulation of IgE-sensitized rat basophilic leukemia RBL-2H3 cells induced activation of c-Jun N-terminal kinase (JNK) within a few minutes with maximum activity attained 40 min later. The increase in JNK activity was accompanied with an increase in phosphorylation of c-Jun in the cells. The Ag-induced JNK activation was inhibited by the phosphatidylinositol 3-kinase inhibitors wortmannin (10-100 nM) and LY 294002 (100 microM) but not by the protein kinase C inhibitors calphostin C (1 and 3 microM) and Ro 31-8425 (1 and 3 microM). Pretreatment with dexamethasone (10 and 100 nM) for 18 h inhibited the Ag-induced increase in JNK activity in a concentration-dependent manner. At least 6 h of preincubation with dexamethasone was necessary to inhibit the Ag-induced JNK activation. The phosphorylation of c-Jun induced by the Ag stimulation was reduced by pretreatment with dexamethasone without reduction of the content of c-Jun protein. The Ag-induced activation of the JNK kinase kinase mitogen-activated protein kinase-extracellular signal-regulated kinase kinase-1 was also inhibited by pretreatment with dexamethasone at 10 and 100 nM. These findings indicate that dexamethasone reduces JNK protein level and inhibits the Ag-induced activation of JNK resulting in the inhibition of c-Jun phosphorylation.