Search

Meteorological and pollutions data are collected daily at monitoring stations of a city. This pollutant-related information can be used to build an early warning system, which provides forecast and also alarms health advice to local inhabitants by medical practicians and local government. In the literature, air quality or pollutant level predictive models using multi-layer perceptrons (MLP) have been employed at a variety of cities by environmental researchers. The practical applications of these models however suffer from different drawbacks so that good generalization may not be obtained. Least squares support vector machines (LS-SVM), a novel type of machine learning technique based on statistical learning theory, can be used for regression and time series prediction. LS-SVM can overcome most of the drawbacks of MLP and has been reported to show promising results.

A multiwalled carbon nanotube is introduced in this paper in the formation between the channel and source and drain electrodes of pentacene-based thin-film transistors. Experimental results show that the pentacene TFTs with multiwalled carbon nanotube have superior performance than the Pd/Ti due to reduced contact resistance.

Advances in genome sequencing technology have led to an exploration in the amount of sequence data available, learning from proteins coded for by genomes is a difficult task. Bioinformatics is thus a burgeoning field that holds great promise for deepening our understanding of biochemical pathways, for understanding the genetic differences between species and how they arose, and for understanding the genetic basis of various disease processes. We developed a method for classification and knowledge discovery in membrane and intrinsic unstructured/disordered proteins (IUP). We analyzed the amino acid compositions and biophysical properties of proteins. Our joint transmembrane and IUP predictor utilized biophysical characterizations, feature generation, feature selection and computational intelligence as well as ensemble methods to improve the accuracies and performances

Advent of high-throughput sequencing technology has led to an exploration of DNA sequence data available. Structures and functions of protein sequence coded for by sequenced genomes remain largely unknown. Automated identification of protein functions and interactions have been largely relying on the known 3D structures or sequence homologues. In particular, intrinsic unstructured or disordered proteins lack specific 3D structures and are unconsented during evolution, but play central roles in diseases characterized by protein misfolding and aggregation. Can we assign protein functions to sequences without relying on 3D structures, to provide useful information for the study of diseases? We developed machine learning techniques to rapidly assess protein functions from sequences. The problem of assigning functional classes to proteins is complicated by the fact that a single protein can participate in several different pathways and thus can have multiple functions (due to complex interactions among proteins). It follows that the instances in the resulting classification problem can carry multiple class labels. We have developed a tree-based classifier that capable of classifying multiply-labeled data and gained an insight into the multi-functional nature of proteins. The algorithm has been used with ensemble methods in connection with other computational intelligence to form a committee machine. Results have been compared favorably to those achieved algorithms such as decision trees and support vector machines

In our attempts to construct methods for automated structural prediction and annotation of proteins as well as automated drug design and discovery, the identification of structure and function from the primary structure of a protein is an important, but difficult problem. We extract features using biophysical properties of the different amino acids and using the patterns of poly-peptide sequences. Based on these features we construct different predictors for different tasks. We demonstrate that our classifiers compare favorably to existing classifiers, and we experiment with the use of ensemble methods to enhance our predictors' accuracies and explaining powers. We showed the synergy of approaches from computational intelligence and biophysics is powerful. This work has particular relevance for the study of ion-channels, ligand binding sites, and alternative splicing

A project to replace two existing room-temperature radio frequency (RF) cavities by one CESR-III 500 MHz superconducting radio frequency (SRF) module was initiated for the Taiwan Light Source (TLS) synchrotron ring in 1999. The goals are to double the photon flux of the synchrotron light by doubling the electron beam current and to increase the stability of the electron beam by taking the advantage of the ultra-weak high-order modes (HOM) of the SRF cavity. The SRF module has been routinely operated since February 2005. The NSRRC users have benefited from a very high photon flux stability (Δ I 0 /I 0 ∼ 0.05%) that had never been achieved previously. Here, we report the initial operational experience of the SRF system.

The increasing peak-valley load difference has made it more difficult for voltage regulation in urban transmission networks, which has addressed high requirement for reactive power optimization. The traditional mathematic model of the multi-object reactive power optimization has been modified in this paper, based on which the reactive power flow of the tie lines can be directly controlled. The optimal solutions for the reactive power have been obtained by means of the improved genetic algorithm. The proposed method has been successfully applied in Beijing urban transmission network.

One method of the static voltage stability analysis is to take the singular point of the static power flow Jacobian matrix as the collapse point of the system. This results in a certain error as the system dynamics is neglected. In the small disturbance analysis, the dynamics of generator, excitation system, speed governor and so on were considered, and the point corresponding to the operating point when the eigenvalue of the system state matrix passes through the vertical axis is regarded as the real collapse point of system. This paper makes a comparison for the two methods. As a result the error of the static voltage stability analysis method using Jacobian matrix could be calculated, and the distance between the singular point of the static power flow Jacobian matrix and the collapse point of the system can be clearly analyzed.

Summary form only given, as follows. A two-dimensional three-moment simulation code was developed and performed for the study of the electron cyclotron resonance (ECR) plasma sources based on the self-consistent fluid model which determines the dynamics of the plasma as well as its interactions with the microwave. In particular, the ECR discharges can be characterized by two major parameters, one is the selection of the microwave mode, and the other is the magnetic field configuration. In this study, two different waveguide modes, i.e. TE/sub 01/ and TM/sub 01/, with the same frequency 2.45 GHz, are applied to induce the ECR discharges in the different magnetic field configurations with uniform, diverging, and mirror profiles. The input power and the background argon pressure are scanned in the range of 0.3-1.2 kW and 1-20 mTorr, respectively. In order to examine their effects on the ECR discharges, we diagnose the plasma density profile, plasma generation rate, spatial temperature distribution, and the spatial deposition of the wave power, etc. by investigating these plasma distributions, appropriate operating conditions far the ECR discharge will be suggested.

Recently several AlGaN-InGaN devices such as blue lasers, blue-green LEDs and high frequency transistors have been reported. In past we have demonstrated GaN based intrinsic photoconductors, Schottky barrier and pn-junction detectors with a long-wavelength cutoff at 365 nm. We now report on the fabrication of true solar-blind ultraviolet photoconductive detectors with Al/sub x/Ga/sub 1-x/N active layers over basal plane sapphire substrates. For sensors with aluminum mole fractions ranging from 5 to 61%, the long wavelength cutoff can be varied from 350 to 240 nm. Photoresponsitivities as high as several hundred amperes per watt were measured for sensors with 10 /spl mu/m inter-electrode spacing.

Monolithic integration of a linear array of GaAs double heterojunction LEDs with a silicon CMOS logic circuit has been demonstrated for the first time. Maximum output power of 7.4 mu W was achieved for the LEDs, which have a 30 mu m*30 mu m active area. A fully functional prototype circuit consisting of eight GaAs LEDs driven by a 3-bit CMOS decoder was fabricated. Only a single 5 V supply was required for circuit operation. This prototype demonstrates that monolithic integration of GaAs LEDs with a Si CMOS circuit is feasible for low-cost optical I/O for future computing systems. >

A new physically based Si MOSFET large signal model, BSIM3v3, developed by UC Berkeley, has been evaluated for high-frequency mixed-signal circuit analysis in a frequency domain, harmonic balance simulator. The model is validated using simulated RF power characteristics of automatic load pull measurement at different bias and matching conditions.