Optimal linear combination of neural networks for improvingclassification performance

This paper presents a new method for linearly combining multiple neural network classifiers based on the statistical pattern recognition theory. In our approach, several neural networks are first selected based on which works best for each class in terms of minimizing classification errors. Then, they are linearly combined to form an ideal classifier that exploits the strengths of the individual classifiers. In this approach, the minimum classification error criterion is utilized to estimate the optimal linear weights. In this formulation, because the classification decision rule is incorporated into the cost function, a more suitable better combination of weights for the classification objective could be obtained. Experimental results using artificial and real data sets show that the proposed method can construct a better combined classifier that outperforms the best single classifier in terms of overall classification errors for test data     

Parametric embedding for class visualization

We propose a new method, parametric embedding (PE), that embeds objects with the class structure into a low-dimensional visualization space. PE takes as input a set of class conditional probabilities for given data points and tries to preserve the structure in an embedding space by minimizing a sum of Kullback-Leibler divergences, under the assumption that samples are generated by a gaussian mixture with equal covariances in the embedding space. PE has many potential uses depending on the source of the input data, providing insight into the classifier's behavior in supervised, semisupervised, and unsupervised settings. The PE algorithm has a computational advantage over conventional embedding methods based on pairwise object relations since its complexity scales with the product of the number of objects and the number of classes. We demonstrate PE by visualizing supervised categorization of Web pages, semisupervised categorization of digits, and the relations of words and latent topics found by an unsupervised algorithm, latent Dirichlet allocation.     

Wide-gap layered oxychalcogenide semiconductors: Materials, electronic structures and optoelectronic properties

Applying the concept of materials design for transparent conductive oxides to layered oxychalcogenides, several p-type and n-type layered oxychalcogenides were proposed as wide-gap semiconductors and their basic optical and electrical properties were examined. The layered oxychalcogenides are composed of ionic oxide layers and covalent chalcogenide layers, which bring wide-gap and conductive properties to these materials, respectively. The electronic structures of the materials were examined by normal/inverse photoemission spectroscopy and energy band calculations. The results of the examinations suggested that these materials possess unique features more than simple wide-gap semiconductors. Namely, the layered oxychalcogenides are considered to be extremely thin quantum wells composed of the oxide and chalcogenide layers or 2D chalcogenide crystals/molecules embedded in an oxide matrix. Observation of step-like absorption edges, large band gap energy and large exciton binding energy demonstrated these features originating from 2D density of states and quantum size effects in these layered materials.     

Spectroscopic imaging and the characterization of the autofluorescence properties of human bronchus tissues using UV laser diodes

Ultraviolet laser diodes (UV-LD) were used for the excitation source of autofluorescence (AF) measurements and spectroscopic imaging of the AF originating from the human bronchus was obtained. The AF spectra from normal bronchus tissues were measured and a clear AF spectrum was obtained by using a short wavelength (400 nm) laser diode; the overlap of the AF signal and excitation source could be substantially eliminated. In order to study the origin of AF intensity deterioration from bronchus tissue due to the formation of tumor tissues, the fluorescence spectrum was measured for various AF substances under various conditions. The blue AF signal of elastin and NADH solutions, which could not be easily studied by conventional excitation light sources, as well as the green AF became weak by adding lactic acid. The AF spectrum was measured for 512/spl times/512 pixels and the intensity mapping as a function of emission wavelength was obtained. Two-dimensional information of the AF signal intensity distribution for a certain wavelength component was measured. The feature originating from a region as small as about 100 /spl mu/m could be recognized. Numerical calculations of the data were performed and precise features of the AF were revealed.     

The Atmospheric Scanning Electron Microscope with open sample space observes dynamic phenomena in liquid or gas

Although conventional electron microscopy (EM) requires samples to be in vacuum, most chemical and physical reactions occur in liquid or gas. The Atmospheric Scanning Electron Microscope (ASEM) can observe dynamic phenomena in liquid or gas under atmospheric pressure in real time. An electron-permeable window made of pressure-resistant 100 nm-thick silicon nitride (SiN) film, set into the bottom of the open ASEM sample dish, allows an electron beam to be projected from underneath the sample. A detector positioned below captures backscattered electrons. Using the ASEM, we observed the radiation-induced self-organization process of particles, as well as phenomena accompanying volume change, including evaporation-induced crystallization. Using the electrochemical ASEM dish, we observed tree-like electrochemical depositions on the cathode. In silver nitrate solution, we observed silver depositions near the cathode forming incidental internal voids. The heated ASEM dish allowed observation of patterns of contrast in melting and solidifying solder. Finally, to demonstrate its applicability for monitoring and control of industrial processes, silver paste and solder paste were examined at high throughput. High resolution, imaging speed, flexibility, adaptability, and ease of use facilitate the observation of previously difficult-to-image phenomena, and make the ASEM applicable to various fields.     

Mitochonic Acid 5 Improves Duchenne Muscular Dystrophy and Parkinson’s Disease Model of Caenorhabditis elegans

Mitochonic Acid 5 (MA-5) enhances mitochondrial ATP production, restores fibroblasts from mitochondrial disease patients and extends the lifespan of the disease model “Mitomouse”. Additionally, MA-5 interacts with mitofilin and modulates the mitochondrial inner membrane organizing system (MINOS) in mammalian cultured cells. Here, we used the nematode Caenorhabditis elegans to investigate whether MA-5 improves the Duchenne muscular dystrophy (DMD) model. Firstly, we confirmed the efficient penetration of MA-5 in the mitochondria of C. elegans. MA-5 also alleviated symptoms such as movement decline, muscular tone, mitochondrial fragmentation and Ca²⁺ accumulation of the DMD model. To assess the effect of MA-5 on mitochondria perturbation, we employed a low concentration of rotenone with or without MA-5. MA-5 significantly suppressed rotenone-induced mitochondria reactive oxygen species (ROS) increase, mitochondrial network fragmentation and nuclear destruction in body wall muscles as well as endogenous ATP levels decline. In addition, MA-5 suppressed rotenone-induced degeneration of dopaminergic cephalic (CEP) neurons seen in the Parkinson’s disease (PD) model. Furthermore, the application of MA-5 reduced mitochondrial swelling due to the immt-1 null mutation. These results indicate that MA-5 has broad mitochondrial homing and MINOS stabilizing activity in metazoans and may be a therapeutic agent for these by ameliorating mitochondrial dysfunction in DMD and PD.     

Raw Starch Digestion by Bacillus polymyxa beta-Amylase

Unlike plant ß-amylases, Bacillus polymyxa ß-amylase digests raw starch at a significant rate. The action of this bacterial ß-amylase on raw starch is stimulated greatly by the presence of the debranching enzyme pullulanase.     

Direct ethanol production from barley beta-glucan by sake yeast displaying Aspergillus oryzae beta-glucosidase and endoglucanase

Three beta-glucosidase- and two endoglucanase-encoding genes were cloned from Aspergillus oryzae, and their gene products were displayed on the cell surface of the sake yeast, Saccharomyces cerevisiae GRI-117-UK. GRI-117-UK/pUDB7 displaying beta-glucosidase AO090009000356 showed the highest activity against various substrates and efficiently produced ethanol from cellobiose. On the other hand, GRI-117-UK/pUDCB displaying endoglucanase AO090010000314 efficiently degraded barley beta-glucan to glucose and smaller cellooligosaccharides. GRI-117-UK/pUDB7CB codisplaying both beta-glucosidase AO090009000356 and endoglucanase AO090010000314 was constructed. When direct ethanol fermentation from 20 g/l barley beta-glucan as a model substrate was performed with the codisplaying strain, the ethanol concentration reached 7.94 g/l after 24 h of fermentation. The conversion ratio of ethanol from beta-glucan was 69.6% of the theoretical ethanol concentration produced from 20 g/l barley beta-glucan. These results showed that sake yeast displaying A. oryzae cellulolytic enzymes can be used to produce ethanol from cellulosic materials. Our constructs have higher ethanol production potential than the laboratory constructs previously reported.     

Development of a homogeneous competitive immunoassay for phosphorylated protein antigen based on the enhanced fluorescence resonance energy transfer technology

We describe a homogeneous competitive immunoassay for a phosphorylated protein antigen. The assay takes advantage of the enhanced fluorescence resonance energy transfer (FRET) technology, which has a unique characteristic that the FRET signal is increased by the specific interaction of two fluorolabeled leucine zippers. We chose extracellular signal-regulated kinase (ERK) as a model antigen and constructed two molecular probes in which either anti-phosphorylation site antibody or the antigen peptide was chemically conjugated to the enhanced FRET probes. While these molecular probes indicated sufficient FRET signal without antigen, they displayed a significant change in the fluorescent spectrum by mixing with phosphorylated antigens. With this competitive enhanced FRET immunoassay, a phosphorylated ERK concentration within the range from 15 nM to 250 nM could be determined. Because the assay is very simple, it would be applied to not only in vitro assay but also in vivo detection of protein phosphorylation.