This thesis is divided into two parts. In the first part, a general problem of the robust template decomposition with restricted weights for cellular neural networks (CNNs) implementing an arbitrary Boolean function is investigated. In the second part, some primary study of the novel Wilcoxon learning machines is made. In the first part of the thesis for the robust CNN template design, the geometric margin of a linear classifier with respect to a training data set, a notion borrowed from the machine learning theory, is used to define the robustness of an uncoupled CNN implementing a linearly separable Boolean function. Consequently, the so-called maximal margin classifiers can be devised via support vector machines (SVMs) to provide the most robust template design for uncoupled CNNs implementing linearly separable Boolean functions. Some general properties of robust CNNs with or without restricted weights are discussed. Moreover, all robust CNNs with restricted weights are characterized. For an arbitrarily given Boolean function, we propose an algorithm, which is the generalized version of the well known CFC algorithm, to find a sequence of robust uncoupled CNNs implementing the given Boolean function. In the second part of the thesis, we investigate the novel Wilcoxon learning machines (WLMs). The invention of these learning machines was motivated by the Wilcoxon approach to linear regression problems in statistics. The resulting linear regressors are quits robust against outliers, as is well known in statistics. The Wilcoxon learning machines investigated in this thesis include Wilcoxon Neural Network (WNN), Wilcoxon Generalized Radial Basis Function Network (WGRBFN), Wilcoxon Fuzzy Neural Network (WFNN), and Kernel-based Wilcoxon Regressor (KWR).

Abstract With the processing of globalization and the large increases in international trade and openness, it is important to capture the business cycle correlation with the intimate countries for government to make better policies and keep the economy steady. This study investigated the changes in relationships between the G7 business cycle after the European integration. Choosing 1993 the Europe Union (EU) commencement as the segment, we separated the sample period into 1965:1-1992:4 and 1993:1-2006:4.We adopted kinds of unit root tests to exam if these variables were stationary and the Johansen co-integration analysis to test whether the stationary long-run equilibrium exist or not. With the consideration of long run information, the Vector Error Correction Model (VECM) was applied to study the relationship between the business cycles of United State, EU and the other G7 countries. By Johansen co-integration analysis, we found that the stationary long-run relationship did exist between their industrial productions¡]IP¡^. In addition, the VECM evidence supported the emergence of two cyclically coherent groups -- the Euro-zone and the English-speaking countries -- after the EU commencement in 1993. In conclusion, with the greater correlation of business cycles, the party in office should take account of the business cycle movement of the closed countries more deliberately in this regionalization era.

Since petroleum is a kind of exhaustive resource, it can not be regenerated after being consumed. And petroleum is distributed extremely uneven in the world, more than half of petroleum is distributed in the Middle East area. In the recent years, the oil price was so fluctuating and broke the record again and again. However, the productivity of petroleum in Taiwan is very low and we are a price taker. So it turns to be important that how the oil price affects the economy. According to Economics, high oil price often causes the staginflation. In the purpose of this study we examine the long run relationship between oil price and personal income in Taiwan by cointegration theory. And we find that there indeed exists a negative longrun relationship. In addition, we consider a nonlinear model, Threshold Vector Error Correction Model, to test a threhold effect in the long run relationship between variables. Finally we have a result that there is a threshold cointegrating relationship between the oil price and personal income in Taiwan.

The rapid-eye-movement (REM) sleep is one of the most important parts in overnight sleep. In this study, an automatic REM sleep staging rule is introduced. Compared with the traditional REM detection method, a distinct feature of this method is that it only requires two EOG signals and thus reduces the number of input signal channels significantly. We calculate the correlation coefficient series between two EOG signals. By representing such a series with a VQ coding method, several techniques are proposed to improve the classification rate. Experimental results are given to demonstrate the effectiveness of the proposed approach.

Considering the requirements of the commercial and military applications on amplitude and phase linearity, it is necessary to reduce nonlinearity of the amplifiers. There are several linearization techniques that are used to reduce nonlinearity effects. Feedforward linearization technique is known as one of the best linearization methods due to its superior linearization performance and broadband operation. Vector modulators which allows amplitude and phase modulation simultaneously, is the most important component of a feedforward system. In this thesis, first of all a broadband V/UHF vector modulator designed and implemented. Then a feedforward system is investigated and implemented using the designed vector modulator for V/UHF band.

A Microsoft Windows program is designed to implement a Multi-Language Identification system based on formants estimation and vector quantization classifier with n-Gram and HMM. LPC is used here as an effective method for formants feature extraction of the speakers, and a new method for distance measure of VQ is also proposed.

In this thesis, the decision-making problem of a GenCo¡¦s bids is formulated and solved by a fuzzy approach to strategic games. It is assumed that only supply-side bidding is considered in the spot market and a player hopes to determine his bids to maximize his own expected payoffs. By means of fuzzy membership functions and operations, a multi-criteria method is used to analyze the player¡¦s choices. Decision maker can decide what kind of strategy combination is optimal by using weight vector to represent his subjective attitude and moral features about the structure of multi-objectives. From this process, a decision can be made more easily, and the results are also close to reality which is not easily modeled by probabilistic method in the classical game theory. Numerical examples with a game of two GenCo¡¦s and a price-elastic demand are offered to illustrate the salient features of the proposed approach.

Nasopharyngeal carcinoma is a commonly occuring tumor in Southern China. However, the genetic basis underlying its tumorigenicity is still unclear. In eukaryotic cells, progression of the cell cycle is regulated by interactions of cyclins, cyclin dependent kinase (CDKs) and CDK inhibitors (CDKIs). These cell cycle-regulator proteins play important roles in growth of both normal and tumor cells. Many human tumors exhibit deregulation of one or more genes which involved in regulation of cell cycle progression. p27Kip1, a member of the Cip/Kip family, inhibits both cyclin D-CDK4, and cyclin E-CDK2 complexes and regulates progression of the cell cycle from G1 to S phase. Although p27Kip1 gene mutations are rare in human tumors, low expression of p27Kip1 are observed in several cancers, such as colon, breast and esophagus. In our previous study, p27Kip1 shown lower expression in two NPC cell lines compared with NNE and 293 (HEK). A doxycycline inducible construct, pBIG2r/p27Kip1, included a full length of human p27Kip1 cDNA was transfected into two NPC cell lines. Expression of several cell cycle-related genes were analyzed. By increasing p27Kip1 in NPC cell lines, we found that the G1 phase and the doubling time were lengthened. Protein expression of cyclin E and CDK2 were up-regulated. These data suggest that the overexpression of p27Kip1 might be cause NPC cells to arrest at G1 phase and might lead to apoptosis.

These theses whether examine the property stock investors are overconfident or underconfident. This thesis firstly use VAR model and the empirical result shows that the property stock investors existed overconfidence investment behavior in the bull market, especially in low-price stock. However, this kind of behavior does not exist in the bear market. The result also shows that most of turnover rate are driven by market return rather than property stock return. this represents investors purchase (or sell) the property stock is caused by the market goes up (or down) rather than have perspective in the property stock. The result implied the property stock investor had speculative trading in short term. In long term, investors have insufficient faith in the real estate market. Next, this thesis use EGARCH and the result shows that the overconfidence investors¡¦ excessive trading will increase volatility, and will harm the development of industry. Finally, we discussed the main factor of investors¡¦ underconfidence in property industry, found political environment as well as long recession are the factors.

The thesis proposes a hybrid broadband equivalent model extraction method, and our goal is to combine via structure and irregular transmission line in print circuit board for extraction of broadband SPICE-compatible model by using the time domain algorithm and full wave simulation in frequency domain, respectively. We can construct broadband SPICE-compatible macro-model scalable library with two kind of different extraction methods, tow kind of extraction of equivalent model can construct the circuit structure for designer demand. Every modules of the broadband macro model of the two extraction models are represented by the optimum pole-residue forms. Using a systematic lumped-model extraction technique, all the optimum pole-residue rational functions can be transfered into a corresponding lumped circuit model. The accuracy of Extraction of Broadband Equivalent Models is demonstrated in frequency -domain responses compared with the 3D-FDTD or HFSS simulation. In addition, the extraction model can simulate in commercial tools effectively, ex: Hspice¡BADS. Even the model can simulate signal integrality and power integrality in Hspice or ADS.