• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 1994.04a
• /
• pp.147-147
• /
• 1994

• Journal of Broadcast Engineering
• /
• v.26 no.2
• /
• pp.184-196
• /
• 2021

This paper proposes statistics adaptive linear regression-based object size prediction method for object detection. YOLOv2 and YOLOv3, which are typical deep learning-based object detection algorithms, designed the last layer of a network using statistics adaptive exponential regression model to predict the size of objects. However, an exponential regression model can propagate a high derivative of a loss function into all parameters in a network because of the property of an exponential function. We propose statistics adaptive linear regression layer to ease the gradient exploding problem of the exponential regression model. The proposed statistics adaptive linear regression model is used in the last layer of the network to predict the size of objects with statistics estimated from training dataset. We newly designed the network based on the YOLOv3tiny and it shows the higher performance compared to YOLOv3 tiny on the UFPR-ALPR dataset.

• Magazine of the Korean Society of Agricultural Engineers
• /
• v.22 no.3
• /
• pp.75-87
• /
• 1980

Most hydro]ogic phenomena are the complex and organic products of multiple causations like climatic and hydro-geological factors. A certain significant correlation on the run-off in river basin would be expected and foreseen in advance, and the effect of each these causual and associated factors (independant variables; present-month rainfall, previous-month run-off, evapotranspiration and relative humidity etc.) upon present-month run-off(dependent variable) may be determined by multiple regression analysis. Functions between independant and dependant variables should be treated repeatedly until satisfactory and optimal combination of independant variables can be obtained. Reliability of the estimated function should be tested according to the result of statistical criterion such as analysis of variance, coefficient of determination and significance-test of regression coefficients before first estimated multiple regression model in historical sequence is determined. But some error between observed and estimated run-off is still there. The error arises because the model used is an inadequate description of the system and because the data constituting the record represent only a sample from a population of monthly discharge observation, so that estimates of model parameter will be subject to sampling errors. Since this error which is a deviation from multiple regression plane cannot be explained by first estimated multiple regression equation, it can be considered as a random error governed by law of chance in nature. This unexplained variance by multiple regression equation can be solved by stochastic approach, that is, random error can be stochastically simulated by multiplying random normal variate to standard error of estimate. Finally hybrid model on estimation of monthly run-off in nonhistorical sequence can be determined by combining the determistic component of multiple regression equation and the stochastic component of random errors. Monthly run-off in Naju station in Yong-San river basin is estimated by multiple regression model and hybrid model. And some comparisons between observed and estimated run-off and between multiple regression model and already-existing estimation methods such as Gajiyama formula, tank model and Thomas-Fiering model are done. The results are as follows. (1) The optimal function to estimate monthly run-off in historical sequence is multiple linear regression equation in overall-month unit, that is; Qn=0.788Pn+0.130Qn-1-0.273En-0.1 About 85% of total variance of monthly runoff can be explained by multiple linear regression equation and its coefficient of determination (R2) is 0.843. This means we can estimate monthly runoff in historical sequence highly significantly with short data of observation by above mentioned equation. (2) The optimal function to estimate monthly runoff in nonhistorical sequence is hybrid model combined with multiple linear regression equation in overall-month unit and stochastic component, that is; Qn=0. 788Pn+0. l30Qn-1-0. 273En-0. 10+Sy.t The rest 15% of unexplained variance of monthly runoff can be explained by addition of stochastic process and a bit more reliable results of statistical characteristics of monthly runoff in non-historical sequence are derived. This estimated monthly runoff in non-historical sequence shows up the extraordinary value (maximum, minimum value) which is not appeared in the observed runoff as a random component. (3) "Frequency best fit coefficient" (R2f) of multiple linear regression equation is 0.847 which is the same value as Gaijyama's one. This implies that multiple linear regression equation and Gajiyama formula are theoretically rather reasonable functions.

• Journal of the Korean Data and Information Science Society
• /
• v.23 no.6
• /
• pp.1231-1239
• /
• 2012

An estimating procedure is introduced for kernel Poisson regression when the input variables consist of numerical and categorical variables, which is based on the penalized negative log-likelihood and the component-wise product of two different types of kernel functions. The proposed procedure provides the estimates of the mean function of the response variables, where the canonical parameter is linearly and/or nonlinearly related to the input variables. Experimental results are then presented which indicate the performance of the proposed kernel Poisson regression.

• Communications for Statistical Applications and Methods
• /
• v.31 no.1
• /
• pp.155-178
• /
• 2024

Regression discontinuity (RD) design is one of the most widely used methods in causal inference for estimation of treatment effect when the treatment is created by a cutpoint from the covariate of interest. There has been little attention to RD design, although it provides a very useful tool for analysis of treatment effect for censored data. In this paper, we define the causal effect for survival function in RD design when the treatment is assigned deterministically by the covariate of interest. We propose estimators of this causal effect for survival data by using transformation, which leads unbiased estimator of the survival function with local linear regression. Simulation studies show the validity of our approach. We also illustrate our proposed method using the prostate, lung, colorectal and ovarian (PLCO) dataset.

• Journal of the Korean Data and Information Science Society
• /
• v.21 no.2
• /
• pp.309-316
• /
• 2010

Support vector quantile regression (SVQR) is capable of providing more complete description of the linear and nonlinear relationships among response and input variables. In this paper we propose a weighted SVQR for the longitudinal data. Furthermore, we introduce the generalized approximate cross validation function to select the hyperparameters which affect the performance of SVQR. Experimental results are the presented, which illustrate the performance of the proposed SVQR.

• Journal of the Korean Data and Information Science Society
• /
• v.18 no.3
• /
• pp.767-772
• /
• 2007

A kernel machine is proposed as an estimating procedure for the linear and nonlinear Poisson regression, which is based on the penalized negative log-likelihood. The proposed kernel machine provides the estimate of the mean function of the response variable, where the canonical parameter is related to the input vector in a nonlinear form. The generalized cross validation(GCV) function of MSE-type is introduced to determine hyperparameters which affect the performance of the machine. Experimental results are then presented which indicate the performance of the proposed machine.

• Journal of the Korean Data and Information Science Society
• /
• v.27 no.6
• /
• pp.1653-1660
• /
• 2016

Smoothly clipped absolute deviation (SCAD) penalty is known to satisfy the desirable properties for penalty functions like as unbiasedness, sparsity and continuity. In this paper, we deal with the regression function estimation and variable selection based on SCAD penalized censored regression model. We use the local linear approximation and the iteratively reweighted least squares algorithm to solve SCAD penalized log likelihood function. The proposed method provides an efficient method for variable selection and regression function estimation. The generalized cross validation function is presented for the model selection. Applications of the proposed method are illustrated through the simulated and a real example.

• Communications for Statistical Applications and Methods
• /
• v.6 no.3
• /
• pp.849-860
• /
• 1999

This paper addresses the issue of estimating regression function with errors in variables using wavelets. We adopt a nonparametric approach in assuming that the regression function has no specific parametric form, To account for errors in covariates deconvolution is involved in the construction of a new class of linear wavelet estimators. using the wavelet characterization of Besov spaces the question of regression estimation with Besov constraint can be reduced to a problem in a space of sequences. Rates of convergence are studied over Besov function classes $B_{spq}$ using $L_2$ error measure. It is shown that the rates of convergence depend on the smoothness s of the regression function and the decay rate of characteristic function of the contaminating error.

• Communications for Statistical Applications and Methods
• /
• v.9 no.1
• /
• pp.101-113
• /
• 2002

In this paper we introduce a new score generating (unction for the rank regression in the linear regression model. The score function compares the $\gamma$'th and s\`th power of the tail probabilities of the underlying probability distribution. We show that the rank estimate asymptotically converges to a multivariate normal. further we derive the asymptotic Pitman relative efficiencies and the most efficient values of $\gamma$ and s under the symmetric distribution such as uniform, normal, cauchy and double exponential distributions and the asymmetric distribution such as exponential and lognormal distributions respectively.