Browsing by Author "Cheng, CH"

Now showing 1 - 9 of 9

A learnable cellular neural network structure with ratio memory for image processing
(2002-12-01)
Wu, CY
;
Cheng, CH
;
電子工程學系及電子研究所
;
Department of Electronics Engineering and Institute of Electronics
In this paper, a learnable cellular neural network (CNN) with space-variant templates and ratio memory (RM) called the RMCNN, is proposed and analyzed. By incorporating both modified Hebbian learning rule and RM into CNN architecture, the RMCNN as the associative memory can generate the absolute weights and then transform them into the ratioed A-template weights as the ratio memories for recognition of noisy input patterns. It is found from simulation results that due to the feature enhancement effect of RM, the RMCNN under constant leakage on template coefficients can store and recognize more patterns than the CNN associative memories without RM, but with the same learning rule and the same constant leakage on space-variant template coefficients. For 9 x 9 (18 x 18) RMCNNs, three (five) patterns can be learned, stored and recognized. Based upon the RMCNN architecture, an experimental chip of CMOS 9 x 9 RMCNN is designed and fabricated by using 0.35 mum CMOS technology. The measurement results have successfully verified the correct functions of RMCNN.
A new structure of large-neighborhood cellular nonlinear network (LN-CNN)
(2003-01-01)
Cheng, CH
;
Chen, SH
;
Lin, LJ
;
Huang, KH
;
Wu, CY
In this paper, a novel Large-Neighborhood Cellular Nonlinear Network (LN-CNN) structure is proposed and analyzed. The proposed LN-CNN structure can realize both A and B templates with more than two neighborhood layers without complex direct connections between neural cell and neighboring cells. In both A and B templates, the first layer is defined by 4 neighboring cells located at the 4 corners of a diamond shape whereas the second layer is defined by 8 cells. In realizing the 12 template coefficients of a template, only 8 connections are required as compared to 12 connections in the conventional CNN structure. Thus the required chip area for synaptic connection can be reduced. Using the proposed LN-CNN structure, the LN-CNN functions, such as noise removing, Muller-Layer arrowhead illusion, and connected component detection, have been successfully realized and verified in MATLAB simulations. The constraints on the realized templates and the template coefficients in the third or higher layers are analyzed and discussed. Based upon he above successful simulation results, the application of the proposed LN-CNN structure to the design of LN-CNN Universal Machine (LN-CNNUM) is quite feasible. The related research will be conducted in the future.
A new two-layer quantum-dot large-neighborhood cellular nonlinear network (QLN-CNN) using quantum-dot cellular automata
(2002)
Tsai, TC
;
Sun, M
;
Lin, LJ
;
Huang, KH
;
Cheng, CH
A novel structure of QLN-CNN is proposed. The new structure of the QLN-CNN contains two layers of QCA array and introduces concepts of the A cloning templates, the B cloning templates, and the thresholds. The gene-decoding concept of CNNs is also used in this structure. Thus it operates more like real CNNs. In this paper, three CNN operations are proven to be achieved by this new structure through MATLAB simulations.
Dynamic fuzzy OWA model for group multiple criteria decision making
(2006-05-01)
Chang, JR
;
Ho, TH
;
Cheng, CH
;
Chen, AP
;
資訊管理與財務金融系註：原資管所+財金所
Obtaining relative weights in MCDM problems is a very important issue. The Ordered Weighted Averaging (OWA) aggregation operators have been extensively adopted to assign the relative weights of numerous criteria. However, previous aggregation operators ( including OWA) are independent of aggregation situations. To solve the problem, this study proposes a new aggregation model-dynamic fuzzy OWA based on situation model, which can modify the associated dynamic weight based on the aggregation situation and can work like a "magnifying lens" to enlarge the most important attribute dependent on minimal information, or can obtain equal attribute weights based on maximal information. Two examples are adopted in this paper for comparison and showing the effects under different weights.
Evaluating the airline service quality by fuzzy OWA operators
(2005)
Cheng, CH
;
Chang, JR
;
Ho, TH
;
Chen, AP
;
資訊管理與財務金融系註：原資管所+財金所
The OWA (Ordered Weighted Averaging) aggregation operators have been extensively adopted to assign the relative weights of numerous criteria. However, previous aggregation operators (including OWA) are independent of aggregation situations. To solve the problem, this study proposes a new aggregation model - dynamic fuzzy OWA operators based on situation model, which can modify the associated dynamic weight based on the aggregation situation and can work like a "magnifying lens" to enlarge the most important attribute dependent on minimal information, or can obtain equal attribute weights based on maximal information. We also apply proposed model to evaluate the service quality of airline.
In the blink of a silicon eye
(2001-05-01)
Cheng, CH
;
Wu, CY
;
Sheu, B
;
Lin, LJ
;
Huang, KH
The design of cellular neural network with ratio memory for pattern learning and recognition
(2000)
Wu, CY
;
Cheng, CH
;
電子工程學系及電子研究所
;
Department of Electronics Engineering and Institute of Electronics
In this paper the cellular neural network (CNN) with ratio memory (RM) is implemented in CMOS to recognize and classify the image patterns. In the implemented CMOS CNN, the BJT-based combined four-quadrant multiplier and two-quadrant divider with separated magnitude and sign is used to implement the Hebbien learning function and the ratio memory. Thus, the combined multiplier and divider and the CNN have simple structure and large input/output signal range. The pattern learning and recognition function of the 9x9 CNN with RM is simulated by both Matlab software and HSPICE. II has been verified that the CNN with RM has the advantages of move scored patterns for processing, and longer memory time with feature enhancement as compared to the CNN without RM.. Thus the proposed CNN with RM has great potential in the applications of neural associate memory for image processing.
The design of ratio-memory cellular neural network (RMCNN) with self-feedback template weight for pattern learning and recognition
(2002)
Cheng, CH
;
Wu, CY
;
電子工程學系及電子研究所
;
Department of Electronics Engineering and Institute of Electronics
In this paper, a new type of the ratio-memory cellular neural network (RMCNN) with spatial-dependent self-feedback A-template weights is proposed and designed to recognize and classify the black-white image patterns. In the proposed RMCNN, the combined four-quadrant multiplier and two-quadrant divider with separated magnitude and sign is used to implement the Hebbien learning function and the ratio memory. To enhance the capability of pattern learning and recognition from noisy input patterns, the Z-ternplate and the spatial-dependent self-feedback weights in the template A are applied to the proposed new type of RMCNN. The pattern learning and recognition function of the 18x18 RMCNN is simulated by Matlab software. It has been verified that the advanced RMCNN has the advantages of more stored patterns for recognition, and better recovery rate as compared to the original RMCNN. Thus the proposed RMCNN has great potential in the applications of neural associate memory for image processing.
Tuning the emission and morphology of cyclometalated iridium complexes and their applications to organic light-emitting diodes
(2005)
Wu, FI
;
Su, HJ
;
Shu, CF
;
Luo, LY
;
Diau, WG
We have synthesized two highly efficient phosphorescent iridium metal complexes, Ir( DPQ)(2)( acac) and Ir(FPQ)(2)(acac), which are based on cyclometalated quinoline ligands, and discuss details of their electrochemical behavior and photophysical properties (viz. absorption and photoluminescence). Single-crystal X-ray diffraction studies of Ir(DPQ)(2)(acac) reveal a distorted octahedral geometry, in which the quinoline N atoms and the C atoms of the orthometalated phenyl groups are located at mutual trans and cis positions, respectively. In contrast, Ir(FPQ)(2)( acac) is an amorphous solid and undergoes a glass transition at 92 degreesC, which we attribute to the presence of the long di-n-octyl chains in the fluorenyl groups. The phosphorescence of these Ir complexes originates from the dominant 3 MLCT excited state shifts to red that occur upon introducing a phenyl substituent and/or a large conjugating aromatic ring into the ligand. A polymer light-emitting diode (PLED) device that uses Ir(FPQ)(2)(acac) as a phosphorescent dopant and a PVK/PBD blend as the host material produces very high efficiency (an external quantum efficiency of 8.16% at 100 mA cm(-2)) and a pure-red emission with 1931 CIE ( Commission Internationale de L'Eclairage) chromaticity coordinates of (x = 0.68, y = 0.32).