• The Journal of the Acoustical Society of Korea
• /
• v.38 no.5
• /
• pp.607-613
• /
• 2019

This paper proposes LF-MMI (Lattice Free Maximum Mutual Information)-based acoustic modeling using external knowledge for speech recognition. Note that an external knowledge refers to text data other than training data used in acoustic model. LF-MMI, objective function for optimization of training DNN (Deep Neural Network), has high performances in discriminative training. In LF-MMI, a phoneme probability as prior probability is used for predicting posterior probability of the DNN-based acoustic model. We propose using external knowledges for training the prior probability model to improve acoustic model based on DNN. It is measured to relative improvement 14 % as compared with the conventional LF-MMI-based model.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.28 no.2A
• /
• pp.197-205
• /
• 2008

In order to estimate mean wind speed in the Gwangyang-Myodo Region, the probability distribution model of extreme values has been used in the statistical analysis of joint distribution probability of daily maximum wind speed and corresponding direction in this paper. For this purpose frequency of daily maximum records at respective stations is inquired into and sample of largest yearly wind speed of sixteen compass direction and non-direction is extracted from daily data of maximum wind speed and appropriate direction of the meteorological observing stations nearby the bridge construction site. These extreme speed records are applied to Gumbel and Weibull distribution model and parameters are estimated through method of moment and method of least squares etc. And also, distribution and parameters are inquired into whether it is fitted through the probability plot correlation coefficient examination. From fitted parameters the largest yearly wind speed of sixteen compass direction and non-direction is extrapolated taking into account factors regarding sample size of data and distance from the bridge construction site according to the appropriate stations.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.9 no.5
• /
• pp.163-168
• /
• 2009

Most image segmentation methods are to represent observed feature vectors at each pixel, which are assumed as appropriated probability models. These models can be used by statistical estimating or likelihood clustering algorithms of feature vectors. EM algorithms have some calculation problems of maximum likelihood for unknown parameters from incomplete data and maximum value in post probability distribution. First, the performance is dependent upon starting positions and likelihood functions are converged on local maximum values. To solve these problems, we mixed the Gausian function and histogram at all the level values at the image, which are proposed most suitable image segmentation methods. This proposed algoritms are confirmed to classify most edges clearly and variously, which are implemented to MFC programs.

• Nuclear Engineering and Technology
• /
• v.21 no.3
• /
• pp.193-204
• /
• 1989

This paper develops a new method for reconstructing neutron flux distribution, that is based on the maximum entropy Principle in information theory. The Probability distribution that maximizes the entropy Provides the most unbiased objective Probability distribution within the known partial information. The partial information are the assembly volume-averaged neutron flux, the surface-averaged neutron fluxes and the surface-averaged neutron currents, that are the results of the nodal calculation. The flux distribution on the boundary of a fuel assembly, which is the boundary condition for the neutron diffusion equation, is transformed into the probability distribution in the entropy expression. The most objective boundary flux distribution is deduced using the results of the nodal calculation by the maximum entropy method. This boundary flux distribution is then used as the boundary condition in a procedure of the imbedded heterogeneous assembly calculation to provide detailed flux distribution. The results of the new method applied to several PWR benchmark problem assemblies show that the reconstruction errors are comparable with those of the form function methods in inner region of the assembly while they are relatively large near the boundary of the assembly. The incorporation of the surface-averaged neutron currents in the constraint information (that is not done in the present study) should provide better results.

• The Korean Journal of Applied Statistics
• /
• v.20 no.3
• /
• pp.573-584
• /
• 2007

A well estimated probability of default is most important for constructing a good credit scoring process. The maximum expected utility (MEU) model has been suggested as an alternative of the traditional logistic regression model. Because the MEU model has been constructed using financial data arising from North America and European countries, the MEU model may not be suitable to Korean private firms. Thus, we propose a Korea-specific MEU model by estimating the parameters involved in kernel functions. This Korea-specific MEU model is illustrated using 34,057 private firms to show the performance of the MEU model relative to the usual logistic regression model.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.28 no.5B
• /
• pp.515-524
• /
• 2008

An estimation of reliable probability precipitation is one of the most important processes for reasonable hydrologic structure design. A probability precipitation has been calculated by frequency analysis using annual maximum rainfall series on the each duration among the observed rainfall data. Annual maximum rainfall series have abstracted on hourly rainfall data or daily rainfall data. So, there is necessary to proper conversion factor between the fixed and sliding durations. Therefore, in this study, conversion factors on the each duration between fixed and sliding durations have calculated using minutely data compared to hourly and daily data of 37 stations observed by Meteorological Administration in Korea. Also, regression equations were computed by regression analysis of conversion factors on the each duration. Consequently, conversion factors were used basis data for calculations of stable probability precipitation.

• Journal of Korea Water Resources Association
• /
• v.32 no.3
• /
• pp.247-254
• /
• 1999

The frequency analyses of annual maximum rainfall data for 22 rainfall gauging stations is Korea were performed. The method of moments (MOM), maximum likelihood (ML), and probability weighted moments (PWM) were used in parameter estimation. The GEV distribution was selected as an appropriate model for annual maximum rainfall data based on parameter validity condition, graphical analysis, separation effect, and goodness of fit tests. For the selected GEV model, spatial analysis was performed and rainfall intensity-duration-frequency equation was derived by using linearization technique. The derived rainfall intensity-duration-frequency equation can be used for estimating rainfall quantiles of the selected stations with convenience and reliability in practice.

• Korean Journal of Agricultural and Forest Meteorology
• /
• v.19 no.3
• /
• pp.130-139
• /
• 2017

The crop damage caused by strong wind was predicted using the wind speed data available from Korean Meteorological Administration (KMA). Wind speed data measured at 19 automatic weather stations in 2012 were compared with wind data available from the KMA's digital forecast. Linear regression equations were derived using the maximum value of wind speed measurements for the three-hour period prior to a given hour and the digital forecasts at the three-hour interval. Estimates of daily maximum wind speed were obtained from the regression equation finding the greatest value among the maximum wind speed at the three-hour interval. The estimation error for the daily maximum wind speed was expressed using normal distribution and Weibull distribution probability density function. The daily maximum wind speed was compared with the critical wind speed that could cause crop damage to determine the level of stages for wind damage, e.g., "watch" or "warning." Spatial interpolation of the regression coefficient for the maximum wind speed, the standard deviation of the estimation error at the automated weather stations, the parameters of Weibull distribution was performed. These interpolated values at the four synoptic weather stations including Suncheon, Namwon, Imsil, and Jangsu were used to estimate the daily maximum wind speed in 2012. The wind damage risk was determined using the critical wind speed of 10m/s under the assumption that the fruit of a pear variety Mansamgil would begin to drop at 10 m/s. The results indicated that the Weibull distribution was more effective than the normal distribution for the estimation error probability distribution for assessing wind damage risk.

• Journal of Digital Convergence
• /
• v.15 no.11
• /
• pp.107-112
• /
• 2017

This paper performed an experimental study to test the validity of the prospect theory proposed by Tversky and Kahneman as an alternative to the expected utility theory. 115 college students attended the hypothetical games to choose one of two lotteries, one is safe option while the other one is risky. The risky options were set up to have low, medium or high probability of payoffs or losses. The amount of payoffs and losses of the lotteries was either large or small. Maximum likelihood estimation of the hypothetical games have shown that in case of high probability of positive payoffs the respondents were risk averse and when the probability of positive payoffs were small the respondents were risk loving. when the possibility of loss is high they were risk loving, while the probability is of loss is low the respondents were found to be risk averse. When the probability of risky options were medium the results were significant statistically in case of only losses. The amount of positive payoff or losses does not affect the results. Overall the results of this experiments support the prospect theory more than those of Laury & Holts (2008).

• Journal of Internet Computing and Services
• /
• v.9 no.5
• /
• pp.141-151
• /
• 2008

In the paper, we analyze the performance of random access channels in multi-service multi-user OFDMA systems. The resource of the random access channels in OFDMA systems are available sub-channels and PN-codes. For given available sub-channels and PN-codes,' we analyze the performance of the random access channels of OFDMA systems in terms of the access success probability, the blocking probability, the access delay and the throughput per each service class. Further, we find the feasible region of the access probability of each service class in which the allowable minimum access success probability, the allowable maximum blocking probability and the allowable maximum access delay are satisfied. The results also can be used to find proper region of the access probabilities of each service class for differentiated quality of service(QoS)s, and for the system operations in multi-service multi-user OFDMA systems.