Sooyoung Jang

Prof. Dr. Sooyoung Jang

Department of Computer Engineering, Hanbat National University, Daejeon, Republic of Korea
orcid

Research Focus & Metrics

Academic Career Length

15Years (2012-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
4 August 2022 in Sensors
Verified
Journal Article
Open access
PDF file
SJSooyoung Jang
Sooyoung Jang
HKHyung-Il Kim
Hyung-Il Kim
Effective exploration is one of the critical factors affecting performance in deep reinforcement learning. Agents acquire data to learn the optimal policy through exploration, and if it is not guaranteed, the data quality deteriorates, which leads to performance degradation. This study investigates the effect of initial entropy, which significantly influences exploration, especially in the early learning stage. The results of this study on tasks with discrete action space show that (1) low initial entropy increases the probability of learning failure, (2) the distributions of initial entropy for various tasks are biased towards low values that inhibit exploration, and (3) the initial entropy for discrete action space varies with both the initial weight and task, making it hard to control. We then devise a simple yet powerful learning strategy to deal with these limitations, namely, entropy-aware model initialization. The proposed algorithm aims to provide a model with high initial entropy to a deep reinforcement learning algorithm for effective exploration. Our experiments showed that the devised learning strategy significantly reduces learning failures and enhances performance, stability, and learning speed.

1 October 2019
Verified
Conference Paper
Closed access
SJSooyoung Jang
Sooyoung Jang
YSYoungsung Son
Youngsung Son
Hyperparameter optimization has a considerable impact on system performance. However many researchers use the default hyperparameter configurations provided by the machine learning frameworks they use because it require a lot of effort, time and computing resources. In this paper, we analyze the effect of the activation function and the kernel initializer on the performance of deep reinforcement learning. In detail, we compared the performance of 42 combinations of 7 activation functions and 6 kernel initializers on two different network types, namely, multi-layer perceptron (MLP) and convolutional neural network (CNN). From the results, it has been revealed that not only did the optimal combinations differ but also, even with the same hyperparameter, the performance difference was dramatic depending on the network type. This phenomenon was prominent in the kernel initializer.

21 October 2020
Verified
Conference Paper
Closed access
JKJiyeon Kim
Jiyeon Kim
SJSooyoung Jang
Sooyoung Jang
DJDeockgu Jee
Deockgu Jee
EKEunjin Ko
Eunjin Ko
SCSeung Han Choi
Seung Han Choi
MHMi Kyong Han
Mi Kyong Han
The SmartCity Convergence Service Platform (SCSP) based on the 5G, which supports high volume of data (e.g., real-time 4K media) to detect possible abnormal or dangerous situations in the urban, is leading to the provision of attractive services to meet smart city requirements. The 5G-SCSP data consists of collected service data, statistical analysis data, media-based event-metadata, and open data. The 5G-SCSP aims at the ease of smart city data collection, user-friendly accessible data utilization using an open interface and the flexibility of data collection for 3rd party services. This paper introduces the 5G-SCSP based smart city service approaching and easy usage of the collected city data and platform operation management of the provided service.

3 January 2022 in ICT Express
Verified
Journal Article
Open access
SJSooyoung Jang
Sooyoung Jang
HKHyung-Il Kim
Hyung-Il Kim

1 October 2018
Verified
Conference Paper
Closed access
SJSooyoung Jang
Sooyoung Jang
MHMikyong Han
Mikyong Han
The needs for training agent with high dimensional continuous action spaces will increase as the robot hardware such as robotic arms and humanoid robots are becoming more and more sophisticated. However, it is difficult and time-consuming task. To tackle the problem, we combine reward shaping and curriculum learning. More specifically, the rewards are provided to the agent for every step it takes and the difficulty of the problem gradually increases depending on the agent learning. Both reward function and curriculum are designed to make the agent achieve its objective. The simulation results demonstrate that the proposed scheme outperforms the comparisons.

4 January 2026 in Mathematics
Verified
Journal Article
Open access
PDF file
SJSooyoung Jang
Sooyoung Jang
CCChangbeom Choi
Changbeom Choi
The estimation of correlation matrices in high-dimensional data streams presents a fundamental conflict between computational efficiency and statistical robustness. Moment-based estimators, such as Pearson’s correlation, offer linear O(N) complexity but lack robustness. In contrast, high-breakdown methods like the minimum covariance determinant (MCD) are computationally prohibitive (O(Np2+p3)) for real-time applications. This paper introduces Fast Outlier-Robust Correlation Estimation (FORCE), a streaming algorithm that performs adaptive coordinate-wise trimming using the P2 algorithm for streaming quantile approximation, requiring only O(p) memory independent of stream length. We evaluate FORCE against six baseline algorithms—including exact trimmed methods (TP-Exact, TP-TER) that use O(NlogN) sorting with O(Np) storage—across five benchmark datasets spanning synthetic, financial, medical, and genomic domains. FORCE achieves speedups of approximately 470× over FastMCD and 3.9× over Spearman’s rank correlation. On S&P 500 financial data, coordinate-wise trimmed methods substantially outperform FastMCD: TP-Exact achieves the best RMSE (0.0902), followed by TP-TER (0.0909) and FORCE (0.1186), compared to FastMCD’s 0.1606. This result demonstrates that coordinate-wise trimming better accommodates volatility clustering in financial time series than multivariate outlier exclusion. FORCE achieves 76% of TP-Exact’s accuracy while requiring 104× less memory, enabling robust estimation in true streaming environments where data cannot be retained for batch processing. We validate the 25% breakdown point shared by all IQR-based trimmed methods using the ODDS-satellite benchmark (31.7% contamination), confirming identical degradation for FORCE, TP-Exact, and TP-TER. For memory-constrained streaming applications with contamination below 25%, FORCE provides the only viable path to robust correlation estimation with bounded memory.