Showing 1–10 of 10 results for author: Chang, K H

Dynamical prediction of two meteorological factors using the deep neural network and the long short term memory $(1)$

Authors: Ki Hong Shin, Jae Won Jung, Sung Kyu Seo, Cheol Hwan You, Dong In Lee, Jisun Lee, Ki Ho Chang, Woon Seon Jung, Kyungsik Kim

Abstract: It is important to calculate and analyze temperature and humidity prediction accuracies among quantitative meteorological forecasting. This study manipulates the extant neural network methods to foster the predictive accuracy. To achieve such tasks, we analyze and explore the predictive accuracy and performance in the neural networks using two combined meteorological factors (temperature and humid… ▽ More It is important to calculate and analyze temperature and humidity prediction accuracies among quantitative meteorological forecasting. This study manipulates the extant neural network methods to foster the predictive accuracy. To achieve such tasks, we analyze and explore the predictive accuracy and performance in the neural networks using two combined meteorological factors (temperature and humidity). Simulated studies are performed by applying the artificial neural network (ANN), deep neural network (DNN), extreme learning machine (ELM), long short-term memory (LSTM), and long short-term memory with peephole connections (LSTM-PC) machine learning methods, and the accurate prediction value are compared to that obtained from each other methods. Data are extracted from low frequency time-series of ten metropolitan cities of South Korea from March 2014 to February 2020 to validate our observations. To test the robustness of methods, the error of LSTM is found to outperform that of the other four methods in predictive accuracy. Particularly, as testing results, the temperature prediction of LSTM in summer in Tongyeong has a root mean squared error (RMSE) value of 0.866 lower than that of other neural network methods, while the mean absolute percentage error (MAPE) value of LSTM for humidity prediction is 5.525 in summer in Mokpo, significantly better than other metropolitan cities. △ Less

Submitted 16 January, 2021; originally announced January 2021.

Comments: 33 Pages, 9 figures

Interpretable Active Learning

Authors: Richard L. Phillips, Kyu Hyun Chang, Sorelle A. Friedler

Abstract: Active learning has long been a topic of study in machine learning. However, as increasingly complex and opaque models have become standard practice, the process of active learning, too, has become more opaque. There has been little investigation into interpreting what specific trends and patterns an active learning strategy may be exploring. This work expands on the Local Interpretable Model-agno… ▽ More Active learning has long been a topic of study in machine learning. However, as increasingly complex and opaque models have become standard practice, the process of active learning, too, has become more opaque. There has been little investigation into interpreting what specific trends and patterns an active learning strategy may be exploring. This work expands on the Local Interpretable Model-agnostic Explanations framework (LIME) to provide explanations for active learning recommendations. We demonstrate how LIME can be used to generate locally faithful explanations for an active learning strategy, and how these explanations can be used to understand how different models and datasets explore a problem space over time. In order to quantify the per-subgroup differences in how an active learning strategy queries spatial regions, we introduce a notion of uncertainty bias (based on disparate impact) to measure the discrepancy in the confidence for a model's predictions between one subgroup and another. Using the uncertainty bias measure, we show that our query explanations accurately reflect the subgroup focus of the active learning queries, allowing for an interpretable explanation of what is being learned as points with similar sources of uncertainty have their uncertainty bias resolved. We demonstrate that this technique can be applied to track uncertainty bias over user-defined clusters or automatically generated clusters based on the source of uncertainty. △ Less

Submitted 23 June, 2018; v1 submitted 31 July, 2017; originally announced August 2017.

Comments: 13 pages, 8 figures, presented at 2018 Conference on Fairness, Accountability, and Transparency (FAT*), New York, New York, USA. Proceedings of the 1st Conference on Fairness, Accountability and Transparency, PMLR 81:49-61, 2018

Optimized planning target volume margin in helical tomotherapy for prostate cancer: is there a preferred method?

Authors: Yuan Jie Cao, Suk Lee, Kyung Hwan Chang, Jang Bo Shim, Kwang Hyeon Kim, Min Sun Jang, Won Sup Yoon, Dae Sik Yang, Young Je Park, Chul Yong Kim

Abstract: To compare the dosimetrical differences between plans generated by helical tomotherapy using 2D or 3D margining technique in in prostate cancer. Ten prostate cancer patients were included in this study. For 2D plans, planning target volume (PTV) was created by adding 5 mm (lateral/anterior-posterior) to clinical target volume (CTV). For 3D plans, 5 mm margin was added not only in lateral/anterior-… ▽ More To compare the dosimetrical differences between plans generated by helical tomotherapy using 2D or 3D margining technique in in prostate cancer. Ten prostate cancer patients were included in this study. For 2D plans, planning target volume (PTV) was created by adding 5 mm (lateral/anterior-posterior) to clinical target volume (CTV). For 3D plans, 5 mm margin was added not only in lateral/anterior-posterior, but also in superior-inferior to CTV. Various dosimetrical indices, including the prescription isodose to target volume (PITV) ratio, conformity index (CI), homogeneity index (HI), target coverage index (TCI), modified dose homogeneity index (MHI), conformation number (CN), critical organ scoring index (COSI), and quality factor (QF) were determined to compare the different treatment plans. Differences between 2D and 3D PTV indices were not significant except for CI (p = 0.023). 3D margin plans (11195 MUs) resulted in higher (13.0%) monitor units than 2D margin plans (9728 MUs). There were no significant differences in any OARs between the 2D and 3D plans. Overall, the average 2D plan dose was slightly lower than the 3D plan dose. Compared to the 2D plan, the 3D plan increased average treatment time by 1.5 minutes; however, this difference was not statistically significant (p = 0.082). We confirmed that 2D and 3D margin plans are not significantly different with regard to various dosimetric indices such as PITV, CI, and HI for PTV, and OARs with tomotherapy. △ Less

Submitted 12 April, 2015; originally announced April 2015.

MSC Class: 00A79 ACM Class: A.0

Treatment plan comparison of Linac step and shoot,Tomotherapy, RapidArc, and Proton therapy for prostate cancer using dosimetrical and biological index

Authors: Suk Lee, Yuan Jie Cao, Kyung Hwan Chang, Jang Bo Shim, Kwang Hyeon Kim, Nam Kwon Lee, Young Je Park, Chul Yong Kim, Sam Ju Cho, Sang Hoon Lee, Chul Kee Min, Woo Chul Kim, Kwang Hwan Cho, Hyun Do Huh, Sangwook Lim, Dongho Shin

Abstract: The purpose of this study was to use various dosimetrical indices to determine the best IMRT modality technique for treating patients with prostate cancer. Ten patients with prostate cancer were included in this study. Intensity modulated radiation therapy plans were designed to include different modalities, including the linac step and shoot, Tomotherapy, RapidArc, and Proton systems. Various dos… ▽ More The purpose of this study was to use various dosimetrical indices to determine the best IMRT modality technique for treating patients with prostate cancer. Ten patients with prostate cancer were included in this study. Intensity modulated radiation therapy plans were designed to include different modalities, including the linac step and shoot, Tomotherapy, RapidArc, and Proton systems. Various dosimetrical indices, like the prescription isodose to target volume (PITV) ratio, conformity index (CI), homogeneity index (HI), target coverage index (TCI), modified dose homogeneity index (MHI), conformation number (CN), critical organ scoring index (COSI), and quality factor (QF) were determined to compare the different treatment plans. Biological indices such as the generalized equivalent uniform dose (gEUD), based tumor control probability (TCP), and normal tissue complication probability (NTCP) were also calculated and used to compare the treatment plans. The RapidArc plan attained better PTV coverage, as evidenced by its superior PITV, CI, TCI, MHI, and CN values. Regarding OARs, proton therapy exhibited superior dose sparing for the rectum and bowel in low dose volumes, whereas the Tomotherapy and RapidArc plans achieved better dose sparing in high dose volumes. The QF scores showed no significant difference among these plans (p=0.701). The average TCPs for prostate tumors in the RapidArc, Linac, and Proton plans were higher than the average TCP for Tomotherapy (98.79%, 98.76%, and 98.75% vs. 98.70%, respectively). Regarding the rectum NTCP, RapidArc showed the most favorable result (0.09%), whereas Linac resulted in the best bladder NTCP (0.08%). △ Less

Submitted 11 March, 2015; originally announced March 2015.

Comments: This work was supported by the Korea University Grant

MSC Class: 00A79 ACM Class: A.0

Evolution of the Major Merger Galaxy Pair Fraction at z < 1

Authors: R. C. Keenan, S. Foucaud, R. De Propris, B. C. Hsieh, L. Lin, R. C. Y. Chou, S. Huang, J. H. Lin, K. H. Chang

Abstract: We present a study of the largest available sample of near-infrared selected (i.e., stellar mass selected) dynamically close pairs of galaxies at low redshifts ($z<0.3$). We combine this sample with new estimates of the major-merger pair fraction for stellar mass selected galaxies at $z<0.8$, from the Red Sequence Cluster Survey (RCS1). We construct our low-redshift $K-$band selected sample using… ▽ More We present a study of the largest available sample of near-infrared selected (i.e., stellar mass selected) dynamically close pairs of galaxies at low redshifts ($z<0.3$). We combine this sample with new estimates of the major-merger pair fraction for stellar mass selected galaxies at $z<0.8$, from the Red Sequence Cluster Survey (RCS1). We construct our low-redshift $K-$band selected sample using photometry from the UKIRT Infrared Deep Sky Survey (UKIDSS) and the Two Micron All Sky Survey (2MASS) in the $K-$band ($\sim 2.2~μ$m). Combined with all available spectroscopy, our $K-$band selected sample contains $\sim 250,000$ galaxies and is $> 90\%$ spectroscopically complete. The depth and large volume of this sample allow us to investigate the low-redshift pair fraction and merger rate of galaxies over a wide range in $K-$band luminosity. We find the major-merger pair fraction to be flat at $\sim 2\%$ as a function of $K-$band luminosity for galaxies in the range $10^8 - 10^{12} L_{\odot}$, in contrast to recent results from studies in the local group that find a substantially higher low-mass pair fraction. This low-redshift major-merger pair fraction is $\sim 40-50\%$ higher than previous estimates drawn from $K-$band samples, which were based on 2MASS photometry alone. Combining with the RCS1 sample we find a much flatter evolution ($m = 0.7 \pm 0.1$), in the relation $f_{\rm{pair}} \propto (1+z)^m$, than indicated in many previous studies. These results indicate that a typical $L\sim L^*$ galaxy has undergone $\sim 0.2-0.8$ major mergers since $z=1$ (depending on the assumptions of merger timescale and percentage of pairs that actually merge). △ Less

Submitted 9 October, 2014; v1 submitted 15 August, 2014; originally announced August 2014.

Comments: Revised to match the version accepted to ApJ

Reaction-diffusion processes of three species on small-world networks

Authors: Kyungsik Kim, K. H. Chang, M. -K. Yum, J. S. Choi, T. Odagaki

Abstract: We study the decay process for the reaction-diffusion process of three species on the small-world network. The decay process is manipulated from the deterministic rate equation of three species in the reaction-diffusion system. The particle density and the global reaction rate on a two dimensional small-world network adding new random links is discussed numerically, and the global reaction rate… ▽ More We study the decay process for the reaction-diffusion process of three species on the small-world network. The decay process is manipulated from the deterministic rate equation of three species in the reaction-diffusion system. The particle density and the global reaction rate on a two dimensional small-world network adding new random links is discussed numerically, and the global reaction rate before and after the crossover is also found by means of the Monte Carlo simulation. The time-dependent global reaction rate scales as a power law with the scaling exponent 0.66 at early time regime while it scales with -0.50 at long time regime, in all four cases of the added probability $p=0.2-0.8$. Especially, our result presented is compared with the numerical calculation of regular networks. △ Less

Submitted 29 March, 2005; originally announced March 2005.

Comments: 3 pages, 1figure

Power Law Distributions for Stock Prices in Financial Markets

Authors: Kyungsik Kim, S. -M. Yoon, K. H. Chang

Abstract: We study the rank distribution, the cumulative probability, and the probability density of returns of stock prices of listed firms traded in four stock markets. We find that the rank distribution and the cumulative probability of stock prices traded in are consistent approximately with the Zipf's law or a power law. It is also obtained that the probability density of normalized returns for liste… ▽ More We study the rank distribution, the cumulative probability, and the probability density of returns of stock prices of listed firms traded in four stock markets. We find that the rank distribution and the cumulative probability of stock prices traded in are consistent approximately with the Zipf's law or a power law. It is also obtained that the probability density of normalized returns for listed stocks almost has the form of the exponential function. Our results are compared with those of other numerical calculations. △ Less

Submitted 1 December, 2004; originally announced December 2004.

Comments: 7 pages, 12 figures, latex

Zipf's Law Distributions for Korean Stock Prices

Authors: Kyungsik Kim, S. -M. Yoon, C. Christopher Lee, K. H. Chang

Abstract: This paper investigates the rank distribution, cumulative probability, and probability density of price returns for the stocks traded in the KSE and the KOSDAQ market. This research demonstrates that the rank distribution is consistent approximately with the Zipf's law with exponent $α= -1.00$ (KSE) and -1.31 (KOSDAQ), similar that of stock prices traded on the TSE. In addition, the cumulative p… ▽ More This paper investigates the rank distribution, cumulative probability, and probability density of price returns for the stocks traded in the KSE and the KOSDAQ market. This research demonstrates that the rank distribution is consistent approximately with the Zipf's law with exponent $α= -1.00$ (KSE) and -1.31 (KOSDAQ), similar that of stock prices traded on the TSE. In addition, the cumulative probability distribution follows a power law with scaling exponent $β= -1.23$ (KSE) and -1.45 (KOSDAQ). In particular, the evidence displays that the probability density of normalized price returns for two kinds of assets almost has the form of an exponential function, similar to the result in the TSE and the NYSE. △ Less

Submitted 18 May, 2004; originally announced May 2004.

Comments: 9 pages, 3 figures

Multifractal Measures on Small-World Networks

Authors: Kyungsik Kim, K. H. Chang, S. M. Yoon, C. Christopher Lee, J. S. Choi

Abstract: We investigate the multifractals of the normalized first passage time on one-dimensional small-world network with both reflecting and absorbing barriers. The multifractals is estimated from the distribution of the normalized first passage time charactrized by the random walk on the small-world network with three fractions of edges rewired randomly. Particularly, our estimate is the fractal dimen… ▽ More We investigate the multifractals of the normalized first passage time on one-dimensional small-world network with both reflecting and absorbing barriers. The multifractals is estimated from the distribution of the normalized first passage time charactrized by the random walk on the small-world network with three fractions of edges rewired randomly. Particularly, our estimate is the fractal dimension D_0 = 0.917, 0.926, 0.930 for lattice points L = 80 and a randomly rewired fraction p = 0.2. The numerical result is found to disappear multifractal properties in the regime p> p_c, where p_c is the critical rewired fraction. △ Less

Submitted 6 May, 2004; v1 submitted 6 May, 2004; originally announced May 2004.

Comments: 13 pages, 4 figures

Decay Process for Three - Species Reaction - Diffusion System

Authors: Kyungsik Kim, K. H. Chang, Y. S. Kong

Abstract: We propose the deterministic rate equation of three-species in the reaction - diffusion system. For this case, our purpose is to carry out the decay process in our three-species reaction-diffusion model of the form $A+B+C\to D$. The particle density and the global reaction rate are also shown analytically and numerically on a two-dimensional square lattice with the periodic boundary conditions.… ▽ More We propose the deterministic rate equation of three-species in the reaction - diffusion system. For this case, our purpose is to carry out the decay process in our three-species reaction-diffusion model of the form $A+B+C\to D$. The particle density and the global reaction rate are also shown analytically and numerically on a two-dimensional square lattice with the periodic boundary conditions. Especially, the crossover of the global reaction rate is discussed in both early-time and long-time regimes. △ Less

Submitted 21 December, 1999; v1 submitted 17 December, 1999; originally announced December 1999.

Comments: 6 pages, 3 figures, Latex

Journal ref: J. Phys. Soc. Jpn. Vol.68,1450(1999)