Browsing by Author "Tsai, CJ"

Now showing 1 - 20 of 77

A fuzzy inductive learning algorithm for parallel loop scheduling
(1997)
Tsai, CJ
;
Tseng, SS
;
Wang, CH
;
Yang, CT
;
Jiang, MF
The conventional symbolic learning algorithm can not infer data that contains fuzzy information. In the past few years, we have designed a parallel loop scheduling based upon knowledge based approach that is called KPLS to choose an appropriate schedule for different loop to assign loop iterations to a multiprocessor system for achieving high speedup rates. Unfortunately, we found that these attributes that were applied in KPLS contain some fuzzy information, which are inapplicable to the traditional symbolic learning strategy for inferring some concept descriptions. In this paper, we apply fuzzy set concept to AQR learning algorithm that is called FAQR. FAQR which can induce fuzzy linguistic rules from fuzzy instances is then proposed to solve the above parallel loop scheduling problem. Some promising inference rules have been found and applied to infer the choice of parallel loop scheduling. Besides, we apply FAQR in IRIS Flower Classification Problem. Experimental results show that our method yields high accuracy in both different domains.
A model for the relative humidity effect on the readings of the PM10 beta-gauge monitor
(2003-12-01)
Chang, CT
;
Tsai, CJ
;
環境工程研究所
;
Institute of Environmental Engineering
In the previous paper (Atmos. Environ. 15 (1981) 1087), we found that the PM10 concentrations detected by the Wedding beta-gauge PM10 monitor and those measured by the manual hi-vol PM10 sampler were quite close when the ambient relative humidity (RH) was lower than the deliquescence RH (DRH) of aerosols. However, when the deliquescent point was exceeded, PM10 concentrations of the beta-gauge were found to be higher and differences increased with an increasing ambient RH. In addition, theoretical water mass calculated by a thermodynamic model (ISORROPIA model, (Aquat. Geochem. 4 (1998) 123)) was found to be much higher than the actual values. In this study, models were developed to determine water evaporation loss from collected particles on the filter tape of the beta-gauge during sampling and in the monitoring room. Simulated results show that all absorbed water will evaporate completely at RH lower than about 85%. However, absorbed water does not evaporate completely at RE higher than about 85%, and remaining water in particles accounts for higher beta-gauge readings than the hi-vol concentrations. The simulated daily beta-gauge PM10 concentrations are close to the actual beta-gauge readings obtained previously. (C) 2003 Elsevier Ltd. All rights reserved.
A model-based rate allocation mechanism for wavelet-based embedded image and video coding
(2005)
Yu, YH
;
Tsai, CJ
;
資訊工程學系
;
Department of Computer Science
In wavelet-based embedded coding for still images and/or videos, a trade-off between scalability and coding efficiency is achieved via layered-packetization of the embedded bitstreams optimized for several target operating bitrate points. This process is called rate allocation (tier-2 coding). The typical rate allocation mechanism is formulated as a rate-distortion optimization problem and a simple searching method with slow convergence rate is used. In this paper, a highly efficient model-based rate allocation mechanism is proposed. The algorithm is base on the adaptive analysis of the relationship between rate and distortion of the image/video data. Experiments conducted on a scalable video codec show that the time to obtain the optimal solution is greatly reduced. The techniques can be applied to various wavelet-based embedded image and video coding schemes, such as JPEG2000 and scalable video coding.
A novel visual distortion sensitivity analysis for video encoder bit allocation
(2004)
Tang, CW
;
Chen, CH
;
Yu, YH
;
Tsai, CJ
;
資訊工程學系
A novel video bit allocation technique adopting a visual distortion sensitivity model for better rate-visual distortion coding control is proposed in this paper. Instead of applying complicated semantic understanding, the proposed automatic distortion sensitivity analysis process analyzes both the motion and the texture structures in the video sequences in order to achieve better bit allocation for rate-constrained video coding This analysis evaluates the tolerable perceptual distortions on a macroblock basis, and allocates fewer bits to regions permitting large perceptual distortions for rate reduction. The proposed algorithm can be incorporated into any existing video coding rate control schemes to achieve same Visual quality at greatly reduced bitrate. Experiments based on H.264 show that this technique achieves bit-rate saving of up to 40% with no perceptual quality degradations. The experiments also demonstrate the inadequacy Of using PSNR as a distortion measure in a video coding framework.
A pilot-scale study of the design and operation parameters of a pulse-jet baghouse
(1998-12-01)
Lu, HC
;
Tsai, CJ
;
環境工程研究所
;
Institute of Environmental Engineering
Filtration curves and pulse-jet cleaning performance of different design and operating conditions of a pilot-scale pulse-jet baghouse are investigated. The effective residual pressure loss is used as an index of bag cleaning effect, while the average pulse overpressure inside the bag is used as an index of bag cleaning intensity. It is found that filtration curves vary with the initial tank pressure and the nozzle diameter of the bag cleaning system. The filtration time increases with increasing initial tank pressure or nozzle diameter. However, a critical value of the tank pressure exists for an effective bag cleaning. It is also found that a critical effective residual pressure loss value exists for the pulse-jet cleaning system. Too large an initial tank pressure and nozzle diameter result in a waste of cleaning energy. The addition of a venturi increases the average pulse overpressure appreciably, hence increasing the cleaning effect. A venturi is suggested to be installed to increase the bag cleaning effect and maximize cleaning energy in this study. (C) 1998 American Association for Aerosol Research.
A pilot-scale study of the design and operation parameters of a pulse-jet baghouse
(1998-12-01)
Lu, HC
;
Tsai, CJ
;
環境工程研究所
;
Institute of Environmental Engineering
Filtration curves and pulse-jet cleaning performance of different design and operating conditions of a pilot-scale pulse-jet baghouse are investigated. The effective residual pressure loss is used as an index of bag cleaning effect, while the average pulse overpressure inside the bag is used as an index of bag cleaning intensity. It is found that filtration curves vary with the initial tank pressure and the nozzle diameter of the bag cleaning system. The filtration time increases with increasing initial tank pressure or nozzle diameter. However, a critical value of the tank pressure exists for an effective bag cleaning. It is also found that a critical effective residual pressure loss value exists for the pulse-jet cleaning system. Too large an initial tank pressure and nozzle diameter result in a waste of cleaning energy. The addition of a venturi increases the average pulse overpressure appreciably, hence increasing the cleaning effect. A venturi is suggested to be installed to increase the bag cleaning effect and maximize cleaning energy in this study. (C) 1998 American Association for Aerosol Research.
A two-phase fuzzy mining and learning algorithm for adaptive learning environment
(2001)
Tsai, CJ
;
Tseng, SS
;
Lin, CY
;
資訊工程學系
;
Department of Computer Science
As computer-assisted instruction environment becomes more popular over the world, the analysis of historical learning records of students becomes more important. In this work, we propose a Two-Phase Fuzzy Mining and Learning Algorithm, integrating data mining algorithm, fuzzy set theory, and look ahead mechanism, to find the embedded information, which can be provided to teachers for further analyzing, refining or reorganizing the teaching materials and tests, from historical learning records.
A universal calibration curve for the TSI aerodynamic particle sizer
(2004-05-01)
Tsai, CJ
;
Chen, SC
;
Huang, CH
;
Chen, DR
;
環境工程研究所
A universal calibration curve for the accurate determination of particle aerodynamic diameter by the TSI APS (TSI, Inc., St. Paul, MN, USA) operating at various temperature, pressure, particle, and gas properties is proposed. The previous dimensionless APS response function proposed by Chen et al. (1985) uses the Stokes number as the governing parameter and is valid in the Stokesian regime only. In the non-Stokesian regime, an additional non-Stokesian correction factor is needed to correct for the indicated aerodynamic diameters of the APS. The universal curve in the present study is based on the relationship between V-p* and Stm, where V-p* = V-p /V-g and V-p is particle velocity, V-g is gas velocity at the nozzle exit, and Stm is modified Stokes number. Stm incorporates the non-Stokesian effect and is defined as 24St/(Re* C-D), where Re is the flow Reynolds number and C-D is the particle drag coefficient. We find that the new calibration curve can predict the particle aerodynamic diameter accurately within 6% of error under different operating conditions and particle/gas properties from the referenced condition without the need to introduce additional correction factors.
Adaptive rate-distortion optimization using perceptual hints
(2004)
Tsai, CJ
;
Tang, CW
;
Chen, CH
;
Yu, YH
;
資訊工程學系
A novel video coding approach that performs adaptive rate-distortion optimization guided by perceptual hints is proposed in this paper The key idea is to adaptively adjust the Lagrange multipliers of the RDO coder control module based on visual attention analysis. The observation is that human vision is sensitive to movement of well structured objects while tolerates large distortion in moving areas with random structure (texture-wise and motion-wise). The proposed algorithm analyzes permissible perceptual distortions and, accordingly, assigns larger Lagrange multipliers to regions that are perceptually less sensitive to distortion so that rate-reduction is weighted more than distortion reduction in these regions. Experiments show that this scheme achieves bit-rate saving of 5-7% with virtually same perceptual quality and is very promising for practical systems.
Adsorption capacity of a nylon filter of filter pack system for HCl and HNO3 gases
(2004-02-01)
Tsai, CJ
;
Huang, CH
;
Lu, HH
;
環境工程研究所
;
Institute of Environmental Engineering
Laboratory and field studies were conducted to evaluate the collection capacity of the nylon filter (Nylasorb membrane, 47 mm diameter, 1.0 mum pore size, Gelman Laboratory, USA) for HCl and HNO3. In ambient sampling, the field results show that one piece of the nylon filter in the filter pack of an annular denuder system (ADS) or a honeycomb denuder system (HDS) has enough capacity to adsorb HCl and HNO3 gases evaporated from the particles collected on the Teflon filter. The laboratory results show that the individual capacity of the nylon filter to adsorb HCl gas is 8, 160, and 240 mug and that of HNO3 gas is 160, 1400, 1900 mug at 20% +/- 5%, 55% +/- 5%, and 80% +/- 5% relative humidity, respectively. At low humidity, the efficiency of the nylon filter for HCl gas will be decreased substantially when HNO3 co-exists. It is decreased to as little as 35-50% when relative humidity is low, at 20% +/- 5% for the added amount of HNO3 is from 40 to 145 mug. A monolayer adsorption theory was successfully employed to estimate the reduction of adsorbing capacity to adsorb HCl gas when HNO3 gas co-exists. There is no significant effect on the adsorbing efficiency of HNO3 when HCl co-exists; the collection efficiency is greater than 95% at all relative humidities.
An efficient dual-interpolator architecture for subpixel motion estimation
(2005)
Wang, YY
;
Tsai, CJ
;
資訊工程學系
;
Department of Computer Science
New generation of video codecs typically adopts a sophisticated interpolation filter for sub-pixel motion estimation (ME). For embedded applications, it is crucial to reduce both the memory requirement and the computational complexity for sub-pixel ME. The key observation in this paper is that the interpolation filter for motion estimation and the filter for coding do not have to be the same. By adopting a simpler on-the-fly interpolation filter during ME process and a standard-compliant filter for coding purpose, both memory and complexity can be reduced with very little coding performance degradation. Since Hadamard transformed-SAD is often used in high quality codecs as a better matching measure, we further show that the ME interpolation Mer can be combined with Hadamard transform for efficient VLSI implementation. Initial results show very promising performance of the proposed system.
An efficient venturi scrubber system to remove submicron particles in exhaust gas
(2005-03-01)
Tsai, CJ
;
Lin, CH
;
Wang, YM
;
Hunag, CH
;
Li, SN
An efficient venturi scrubber system making use of heterogeneous nucleation and condensational growth of particles was designed and tested to remove fine particles from the exhaust of a local scrubber where residual SiH4 gas was abated and lots of fine SiO2 particles were generated. In front of the venturi scrubber, normal-temperature fine-water mist mixes with high-temperature exhaust gas to cool it to the saturation temperature, allowing submicron particles to grow into micron sizes. The grown particles are then scrubbed efficiently in the venturi scrubber. Test results show that the present venturi scrubber system is effective for removing submicron particles. For SiO2 particles greater than 0.1 mum, the removal efficiency is greater than 80-90%, depending on particle concentration. The corresponding pressure drop is relatively low. For example, the pressure drop of the venturi scrubber is similar to15.4 +/- 2.4 cm H2O when the liquid-to-gas ratio is 1.50 L/m(3). A theoretical calculation has been conducted to simulate particle growth process and the removal efficiency of the venturi scrubber. The theoretical results agree with the experimental data reasonably well when SiO2 particle diameter is greater than 0.1 mum.
An evolutionary and attribute-oriented ensemble classifier
(2006)
Lee, CI
;
Tsai, CJ
;
Ku, CW
;
資訊工程學系
;
Department of Computer Science
In the research area of decision tree, numerous researchers have been focusing on improving the predictive accuracy. However, obvious improvement can hardly be made until the introduction of the ensemble classifier. In this paper, we propose an Evolutionary Attribute-Oriented Ensemble Classifier (EAOEC) to improve the accuracy of sub-classifiers and at the same time maintain the diversity among them. EAOEC uses the idea of evolution to choose proper attribute subset for the building of every sub-classifier. To avoid the huge computation cost for the evolution, EAOEC uses the gini value gained during the construction of a sub-tree as the evolution basis to build the next sub-tree. Eventually, EAOEC classifier uses uniform weight voting to combine all sub-classifiers and experiments show that EAOEC can efficiently improve the predictive accuracy.
An intelligent parallel loop scheduling for parallelizing compilers
(2000-03-01)
Fann, YW
;
Yang, CT
;
Tseng, SS
;
Tsai, CJ
;
資訊工程學系
In this paper we propose a knowledge-based approach to solving loop-scheduling problems. A rule-based system, called IPLS, is developed by combining a repertory grid and an attribute ordering table to construct a knowledge base. IPLS chooses an appropriate scheduling algorithm by inferring some features of loops and assigning parallel loops to multiprocessors to achieve significant speedup. Because more attributes are proposed, the accuracy of selection of an appropriate scheduling method is improved. In addition, the refined IPLS system can automatically adjust the attributes in the knowledge base according to profile information; therefore, IPLS has the capability of feedback learning. The experimental results show that our approach can achieve greater speedup on multiprocessor systems than can others.
An investigation of dust emissions from unpaved surfaces in Taiwan
(2002-11-01)
Tsai, CJ
;
Chang, CT
;
環境工程研究所
;
Institute of Environmental Engineering
One of the critical problems in making an effective strategy to control fugitive road dust is to estimate the emission factor accurately. The paper presents measured results of emission factor at field and in a wind tunnel. For the field study, three unpaved roads in Hsin Chu, Taiwan were tested. For the wind tunnel testing, the tested road dust was also collected from the field. Results of field study indicate that the emission factor of unpaved roads increases with the increasing wind speed, the number of vehicles and the speed of vehicles. It can be predicted as e (g/m(2)day) = 8.72 x 10(-3) x V-w(0.64) X s(0.19) x M-0.05 x V-v(1.32) x N-v(1.11) (V-w, m/s-wind speed, s, %-silt content of road dust; M, %-moisture content of road dust; N-v, #/h-number of vehicles: V-v, km/h-speed of vehicles) by the multi-variable regression technique. The fractions of PM10 and PM2.5, in the TSP reentrained from the unpaved roads are 20.6 +/- 12.9% and 2.3 +/- 1.2%, respectively. The wind tunnel test results show that air acceleration rate and edge effect of the dust surface increase the emission factors considerably. As air acceleration rate increases from 0.1 to 1.5 m/s(2), the emission factor increases linearly from 1.0 x 10(-4) to 7.0 x 10(-4) kg/m(2) s. However, test results show no significant effect of air acceleration and edge effect of dust surface on the threshold wind speed of reentrainment. (C) 2002 Elsevier Science B.V. All rights reserved.
Artifacts of ionic species for hi-vol PM10 and PM10 dichotomous samplers
(1998-05-01)
Tsai, CJ
;
Perng, SN
;
環境工程研究所
;
Institute of Environmental Engineering
This study employs two Annular Denuder Systems (ADS(10) and ADS(2.5)) to quantify sampling artifacts of soluble ionic species determined by the high-volume PM10 (hi-vol PM10) sampler using quartz fiber filters and PM10 dichotomous (Dichot) sampler using Teflon filters. According to the experimental results, the percentage of artifact of a species over the actual species concentrationis +11, +8 and +15% for PM10HV, PM10DICH, PM2.5DICH, respectively, for sulfate species; -16, -21 and -21% for PM10HV, PM10DICH, PM2.5DICH, respectively, for nitrate species; -24, -32 and -54% for PM10HV, PM10DICH, PM2.5DICH, respectively, for for chloride ion species; -17, -21 and -18% for PM10HV, PM10DICH, PM2.5DICH, respectively, for ammonium ion species. For volatile species such as nitrate, ammonium and chloride ions, additional sampling losses due to gas-particle and particle-particle interactions are more important than pure volatilization loss. Amount of positive and negative artifacts is approximately equal and the amount of net artifact does not account for more than 15% of PM10 concentration. (C) 1998 Elsevier Science Ltd. All rights reserved.
Atmospheric lead concentration distribution in Northern Taiwan
(2003-08-01)
Lu, HC
;
Tsai, CJ
;
Hung, IF
;
環境工程研究所
;
Institute of Environmental Engineering
Atmospheric lead concentrations were measured randomly, approximately once per week, at five traffic sites in northern Taiwan from September 1994 to May 1995. Three types,of theoretical distributions, lognormal, Weibull and gamma were selected to fit the frequency distribution of the measured lead concentration. Four goodness-of-fit criteria were used to judge which theoretical distribution is the most appropriate to represent the frequency distributions of atmospheric lead. The results show that atmospheric lead concentrations in total suspended particulates fit the lognormal distribution reasonably well in northern Taiwan. The intervals of fitted theoretical cumulative frequency distributions (CFDs) can successfully contain the measured data when the population mean is estimated with a 95% confidence interval. In addition, atmospheric lead concentration exceeding a critical concentration is also predicted from the fitted theoretical CFDs. (C) 2003 Elsevier Science Ltd. All rights reserved.
Axial flow cyclone for segregation and collection of ultrafine particles: Theoretical and experimental study
(2005-03-01)
Hsu, YD
;
Chein, HM
;
Chen, TM
;
Tsai, CJ
;
環境工程研究所
In this study, an axial flow cyclone was designed, fabricated, and evaluated at different conditions of air flow rates (Q(0)) and low-pressure environments (P), especially for the segregation and collection of ultrafine particles. An evaporation/condensation type of aerosol generation system consisting of tube furnace and mixing chamber was employed to produce test aerosols. The test aerosol was then classified by a differential mobility analyzer (DMA) and number concentration was measured by a condensation nuclei counter (CNC) and an electrometer upstream and downstream of the cyclone, respectively. The s-shaped curve of the collection efficiency in submicron particle size range was obtained to be similar to the traditional cyclone found in the literatures when the particles were largerthan 40 nm at Q(0) = 1.07, 0.455 L(STP)/min, and P= 4.8-500 Torr. The curve was found to be fitted very well by a serniempirical equation described in this paper. Forparticles smaller than 40 nm, however, the collection efficiency was unusually increased as the particle diameter was decreased due to the fact that the diffusion deposition becomes the dominant collection mechanism in the low pressure conditions. A model composed of centrifugal force and diffusion deposition is presented and used to fit the experimental data. The cyclone was demonstrated to separate and collect ultrafine particles effectively in the tested vacuum conditions.
Building a CAL expert system based upon two-phase knowledge acquisition
(2002-04-01)
Tsai, CJ
;
Tseng, SS
;
資訊工程學系
;
Department of Computer Science
With the fast growing and globally accepted e-learning technology, Computer-Assisted Learning (CAL) system, which can provide the individualized materials, becomes a matter of great importance. In this paper, we propose a CAL Expert System (CAL-ES), in which teachers can provide their teaching strategy as guidance to students' learning. Because the knowledge hierarchy and the knowledge itself are both important in CAL domain, and the teaching strategy can be considered as rule-format, Two-phase Knowledge Acquisition is proposed to acquire the knowledge hierarchy and the rule-based knowledge by the cooperation of teachers and knowledge engineers. The prototype of CAL-ES is constructed of the mathematical teaching materials. The CAL-ES is useful for student learning and is easy for teachers maintaining. (C) 2002 Elsevier Science Ltd. Ail rights reserved.
Building an object-oriented and individualized learning environment on the WWW
(1999)
Su, GH
;
Tseng, SS
;
Tsai, CJ
;
Zheng, JR
;
資訊工程學系
As Internet becomes more popular over the world, more and more teachers want to put the teaching materials on the WWW. So, the problem of how to organize these materials appropriately on the web-based learning environments arises. In this paper, we propose a new model, object-oriented (OO) course model, under the object-oriented paradigm for segmenting and transforming all teaching materials into the teaching objects. Based on the OO course model, the generated course framework would be neither static nor invariable; on the contrary, it can be constructed dynamically from all of the teaching objects and individually in accordance with the aptitude of the students. To manage the teaching objects effectively and efficiently, an object-oriented course construction system (OOCCS), which includes an individualized course construction algorithm and a course framework revision algorithm, is also developed. Comparing to the traditional course formation mechanism, our OOCCS has three main advantages: Individuality, Flexibility and Maintainability.