• International Journal of Internet, Broadcasting and Communication
• /
• v.15 no.3
• /
• pp.66-72
• /
• 2023

Little attention appears to have been paid to the relevance of learning a good representation function in solving long tail tasks. Therefore, we propose a new loss function to ensure a good representation is learnt while learning to classify. We call this loss function Triplet Class-Wise Difficulty-Based (TriCDB-CE) Loss. It is a combination of the Triplet Loss and Class-wise Difficulty-Based Cross-Entropy (CDB-CE) Loss. We prove its effectiveness empirically by performing experiments on three benchmark datasets. We find improvement in accuracy after comparing with some baseline methods. For instance, in the CIFAR-10-LT, 7 percentage points (pp) increase relative to the CDB-CE Loss was recorded. There is more room for improvement on Places-LT.

• Journal of the Korean Data and Information Science Society
• /
• v.14 no.3
• /
• pp.631-639
• /
• 2003

Bayes estimates of reliability for the stress-strength system are obtained with respect to LINEX loss function. A reference prior distribution of the reliability is derived and Bayes estimates of the reliability are also obtained. These Bayes estimates are compared with corresponding estimates under squared-error loss function.

• Journal of the Korean Data and Information Science Society
• /
• v.4
• /
• pp.87-95
• /
• 1993

We find a class of admissible Bayes estimator for the mean vector ${\theta}=({\theta}_{1},{\theta}_{2},...,{\theta}_{p}$ of Poisson distribution under a LINEX loss function. The Monte Carlo Simulation is performed to compare the emprical Bayes estimater under the LINEX loss function and weighted squared error loss respectively.

• Journal of Korean Society for Quality Management
• /
• v.25 no.3
• /
• pp.119-130
• /
• 1997

In this paper, we analyze quality loss function(QLF) of Taguchi. Taguchi method of QLF gives more advanced measure process capability than classic capacity(i.e. Cp). We first discuss of QLF and Cp and give one good example of QLF. Because of simplicity of QLF, it is not good fit to a, pp.y in the real field. We suggest interval quantity-loss cost function and total loss cost(TLC) which modify QLF. Also we give one example which can be obtained in the real field.

• Journal of the Korean Statistical Society
• /
• v.23 no.2
• /
• pp.445-461
• /
• 1994

This paper deals with the problem of obtaining some Bayes estimators and Bayesian credible regions of a reliability function for the Rayleigh distribution. Using several priors for a reliability function some Bayes estimators and Bayes credible sets are proposed and studied under squared error loss and Harris loss. Also the performances and behaviors of the proposed Bayes estimators are examined via Monte Carlo simulations and some numericla examples are given.

• Journal of the Korean Statistical Society
• /
• v.32 no.3
• /
• pp.289-298
• /
• 2003

We consider the estimation problem for the scale parameter of the Rayleigh distribution using weighted balanced loss function (WBLF) which reflects both goodness of fit and precision. Under WBLF, we obtain the optimal estimator which creates a kind of balance between Bayesian and non-Bayesian estimation. We also deal with the estimation of R ＝ P(Y < X) when Y and X are two independent but not identically distributed Rayleigh distribution under squared error loss function.

• 제어로봇시스템학회:학술대회논문집
• /
• 2000.10a
• /
• pp.315-315
• /
• 2000

In this paper, we perform the time series prediction using the SVM(Support Vector Machine). We make use of two different loss functions and two different kernel functions; i) Quadratic and $\varepsilon$-insensitive loss function are used; ii) GRBF(Gaussian Radial Basis Function) and ERBF(Exponential Radial Basis Function) are used. Mackey-Glass time series are used for prediction. For both cases, we compare the results by the SVM to those by ANN(Artificial Neural Network) and show the better performance by SVM than that by ANN.

• Journal of Korean Institute of Industrial Engineers
• /
• v.28 no.2
• /
• pp.216-222
• /
• 2002

The ergonomic human model can be considered as a tool for the evaluation of ergonomic factors in vehicle design process. The proper anthropometric data on driver's postures are needed in order to apply a human model to vehicle design. Although studies on driver's posture have been carried out for the last few decades, there are still some problems for the posture data to be applied directly to the human model due to the lack of fitness because such studies were not carried out under the conditions for the human model application. In the traditional researches, the joint angles were evaluated by the categorized data, which are not appropriate for the human model application because it is so extensive that it can not explain the posture evaluation data in detail. And the human models require whole-body posture evaluation data rather than joint evaluation data. In this study a postural evaluation function was developed not by category data but by the concept of the loss function in quality engineering. The loss was defined as the discomfort in driver's posture and measured by the magnitude estimation technique in the experiment using a seating buck. Four loss functions for the each joint - knee, hip, shoulder, and elbow were developed and a whole-body postural evaluation function was constructed by the regression analysis using these loss functions as independent factors. The developed postural evaluation function shows a good prediction power for the driver's posture discomfort in validation test. It is expected that the driver's postural evaluation function based on the loss function can be used in the human model application to the vehicle design process.

• Fashion & Textile Research Journal
• /
• v.11 no.6
• /
• pp.918-924
• /
• 2009

As the population of overweight and obese children is rising rapidly around the world, there are many researches on purchasing and wearing children's clothing and optimal sizes, but researches on obese children are still inadequate. This study was carried out on 192 obese children over 75% in BMI. The purpose of the study was to set up the optimal interval of sizing system using the loss function which would be a guide for obese children for selecting ready to wear of suitable size. Introducing a loss function, which reflects how much the purchasing desire changes according to the difference, we formulate the problem and suggest a procedure to determine the optimal standard sizes minimizing the loss. These results were as follows ; In size chart of top's, 4 sizes had been determined by a loss function, had covered more than 91.1% of all subjects. In size chart of bottom's, 5 sizes had been determined by a loss function, had covered more than 87.0% of all subjects.

• Communications for Statistical Applications and Methods
• /
• v.21 no.3
• /
• pp.261-274
• /
• 2014

In this paper we develop Bayes and empirical Bayes estimators of the finite population mean with the assumption of posterior linearity rather than normality of the superpopulation under the balanced loss function. We compare the performance of the optimal Bayes estimator with ones of the classical sample mean and the usual Bayes estimator under the squared error loss with respect to the posterior expected losses, risks and Bayes risks when the underlying distribution is normal as well as when they are binomial and Poisson.