• The Journal of the Acoustical Society of Korea
• /
• v.40 no.5
• /
• pp.496-502
• /
• 2021

Recently, deep speaker embedding approach has been widely used in text-independent speaker verification, which shows better performance than the traditional i-vector approach. In this work, to improve the deep speaker embedding approach, we propose a novel method called group-based speaker embedding which incorporates group information. We cluster all speakers of the training data into a predefined number of groups in an unsupervised manner, so that a fixed-length group embedding represents the corresponding group. A Group Decision Network (GDN) produces a group weight, and an aggregated group embedding is generated from the weighted sum of the group embeddings and the group weights. Finally, we generate a group-based embedding by adding the aggregated group embedding to the deep speaker embedding. In this way, a speaker embedding can reduce the search space of the speaker identity by incorporating group information, and thereby can flexibly represent a significant number of speakers. We conducted experiments using the VoxCeleb1 database to show that our proposed approach can improve the previous approaches.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.41 no.4
• /
• pp.238-247
• /
• 2018

This paper considers the problem of scheduling loading and unloading operations of a crane in a railway terminal motivated from rail-road container transshipment operations at Uiwang Inland Container Depot (ICD). Unlike previous studies only considering the total handling time of containers, this paper considers a bi-criteria objective of minimizing the weighted sum of the total handling time and tenant service time. The tenant service time is an important criterion in terms of terminal tenants who are private logistics companies in charge of moving containers from/to the terminal using their trucks. In the rail-road container shipment yard, the tenant service time of a tenant can be defined by a time difference between beginning and finishing loading and unloading operations of a crane. Thus, finding a set of sequences and time of the crane operations becomes a crucial decision issue in the problem. The problem is formulated as a nonlinear program which is improved by linearizing a nonlinear constraint in the model. This paper develops a genetic algorithm to solve the problem and performs a case study on the Uiwang ICD terminal. Computational experiment results show that the genetic algorithm shows better performance than commercial optimization solvers. Operational implications in terms of tenants are drawn through sensitivity analyses.

• The Bulletin of The Korean Astronomical Society
• /
• v.44 no.1
• /
• pp.52.3-52.3
• /
• 2019

We investigate a relative contribution from short to long-term flaring rate to predicting M and X-class flare probabilities. In this study, we consider magnetic parameters summarizing distribution and non-potentiality by Solar Dynamics Observatory/Helioseimic and Magnetic Imager and flare list by Geostationary Operational Environmental Satellites. A short-term rate is the number of major flares that occurred in an given active region (AR) within one day before the prediction time. A mid-term rate is a mean flaring rate from the AR appearance day to one day before the prediction time. A long-term rate is a rate determined from a relationship between magnetic parameter values of ARs and their flaring rates from 2010 May to 2015 April. In our model, the predicted rate is given by the combination of weighted three rates satisfying that their sum of the weights is 1. We calculate Brier skill scores (BSSs) for investigating weights of three terms giving the best prediction performance using ARs from 2015 April to 2018 April. The BSS (0.22) of the model with only long-term is higher than that with only short-term or mid-term. When short or mid-term are considered additionally, the BSSs are improved. Our model has the best performance (BSS = 0.29) when all three terms are considered, and their relative contribution from short to long-term rate are 19%, 23%, and 58%, respectively. This model seems to be more effective when predicting active solar ARs having several major flares.

• Journal of Korean Society of Water and Wastewater
• /
• v.35 no.1
• /
• pp.53-61
• /
• 2021

The quantified analysis of damages to wastewater treatment plants by natural disasters is essential to maintain the stability of wastewater treatment systems. However, studies on the quantified analysis of natural disaster effects on wastewater treatment systems are very rare. In this study, a total disaster index (DI) was developed to quantify the various damages to wastewater treatment systems from natural disasters using two statistical methods (i.e., AHP: analytic hierarchy process and PCA: principal component analysis). Typhoons, heavy rain, and earthquakes are considered as three major natural disasters for the development of the DI. A total of 15 input variables from public open-source data (e.g., statistical yearbook of wastewater treatment system, meteorological data and financial status in local governments) were used for the development of a DI for 199 wastewater treatment plants in Korea. The total DI was calculated from the weighted sum of the disaster indices of the three natural disasters (i.e., TI for typhoon, RI for heavy rain, and EI for earthquake). The three disaster indices of each natural disaster were determined from four components, such as possibility of occurrence and expected damages. The relative weights of the four components to calculate the disaster indices (TI, RI and EI) for each of the three natural disasters were also determined from AHP. PCA was used to determine the relative weights of the input variables to calculate the four components. The relative weights of TI, RI and EI to calculate total DI were determined as 0.547, 0.306, and 0.147 respectively.

• Journal of information and communication convergence engineering
• /
• v.20 no.1
• /
• pp.22-30
• /
• 2022

The reinforcement learning algorithm has proven its potential in solving sequential decision-making problems under uncertainties, such as finding paths to route data packets in wireless sensor networks. With reinforcement learning, the computation of the optimum path requires careful definition of the so-called reward function, which is defined as a linear function that aggregates multiple objective functions into a single objective to compute a numerical value (reward) to be maximized. In a typical defined linear reward function, the multiple objectives to be optimized are integrated in the form of a weighted sum with fixed weighting factors for all learning agents. This study proposes a reinforcement learning -based routing protocol for wireless sensor network, where different learning agents prioritize different objective goals by assigning weighting factors to the aggregated objectives of the reward function. We assign appropriate weighting factors to the objectives in the reward function of a sensor node according to its hop-count distance to the sink node. We expect this approach to enhance the effectiveness of multi-objective reinforcement learning for wireless sensor networks with a balanced trade-off among competing parameters. Furthermore, we propose SDN (Software Defined Networks) architecture with multiple controllers for constant network monitoring to allow learning agents to adapt according to the dynamics of the network conditions. Simulation results show that our proposed scheme enhances the performance of wireless sensor network under varied conditions, such as the node density and traffic intensity, with a good trade-off among competing performance metrics.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.7
• /
• pp.93-100
• /
• 2022

In the era of information overload on the Internet, the recommendation system, which is an indispensable function, is a service that recommends products that a user may prefer, and has been successfully provided in various commercial sites. Recently, studies to reflect the rating time of items to improve the performance of collaborative filtering, a representative recommendation technique, are active. The core idea of these studies is to generate the recommendation list by giving an exponentially lower weight to the items rated in the past. However, this has a disadvantage in that a time function is uniformly applied to all items without considering changes in users' preferences according to the characteristics of the items. In this study, we propose a time-aware collaborative filtering technique from a completely different point of view by developing a new similarity measure that integrates the change in similarity values between items over time into a weighted sum. As a result of the experiment, the prediction performance and recommendation performance of the proposed method were significantly superior to the existing representative time aware methods and traditional methods.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.10
• /
• pp.1531-1536
• /
• 2022

This paper considers a problem to find a set of intervals containing all the points for the given n points and m intervals on a line, This is a special case of the set cover problem, well known as an NP-hard problem. As optimization criteria of the problem, there are minimizing the number of intervals to cover the points, maximizing the number of points each of which is covered by exactly one interval, and so on. In this paper, the intervals have weights and the sum of weights of intervals to cover a point is defined as a membership of the point. We will study the problem to find an interval cover minimizing the maximum of memberships of points. Using the dynamic programming method, we provide an O(m²)-time algorithm to improve the time complexity O(nm log n) given in the previous work.

• Communications for Statistical Applications and Methods
• /
• v.29 no.4
• /
• pp.421-439
• /
• 2022

Phase I dose-finding trials are essential in drug development. By finding the maximum tolerated dose (MTD) of a new drug or treatment, a Phase I trial establishes the recommended doses for later-phase testing. The primary toxicity endpoint of interest is often a binary variable, which describes an event of a patient who experiences dose-limiting toxicity. However, there is a growing interest in dose-finding studies regarding non-binary outcomes, defined by either the weighted sum of rates of various toxicity grades or a continuous outcome. Although several novel methods have been proposed in the literature, accessible software is still lacking to implement these methods. This study introduces a newly developed R package, UnifiedDoseFinding, which implements three phase I dose-finding methods with non-binary outcomes (Quasi- and Robust Quasi-CRM designs by Yuan et al. (2007) and Pan et al. (2014), gBOIN design by Mu et al. (2019), and by a method by Ivanova and Kim (2009)). For each of the methods, UnifiedDoseFinding provides corresponding functions that begin with next that determines the dose for the next cohort of patients, select, which selects the MTD defined by the non-binary toxicity endpoint when the trial is completed, and get oc, which obtains the operating characteristics. Three real examples are provided to help practitioners use these methods. The R package UnifiedDoseFinding, which is accessible in R CRAN, provides a user-friendly tool to facilitate the implementation of innovative dose-finding studies with nonbinary outcomes.

• Journal of Korean Society for Quality Management
• /
• v.50 no.4
• /
• pp.863-876
• /
• 2022

Purpose: The purpose of this study is to propose the product quality satisfaction index based on multiple linear regression using customer reviews. Methods: The proposed framework is composed of four steps. First, we collect online reviews and divide it into insight phrases. The insight phrases are classified using product attribute dictionary and sentiment analysis is conducted. Second, the importance of attributes is calculated in consideration of both regression coefficient and frequency. Third, the positive rate is calculated concerning sentiment analysis result. Therefore, the quality satisfaction index is measured by the weighted sum of importance and positive rate in the last step. Results: We conduct a case study using 2-years(2020, 2021) of Samsung TV reviews to confirm the effectiveness of the proposed methodology. As a result, we found that Picture quality is the most crucial attribute in TV evaluation. The importance of Gaming and content has grown up as the positive rate has also increased. Therefore, the overall satisfaction of TV has increased in 2021 compared to 2020. Conclusion: The result of this study shows that the proposed index reveals the customer's mind efficiently and can be explained by the importance and positive rate of each attribute. By using the proposed index, companies are able to improve and the priority of improvement can be determined.

• KNOM Review
• /
• v.23 no.1
• /
• pp.10-17
• /
• 2020

As the threat of Distributed Denial-of-Service attacks that exploit weakly secure IoT devices has spread, research on source-side Denial-of-Service attack detection is being activated to quickly detect the attack and the location of attacker. In addition, a collaborative source-side attack detection technique that shares detection results of source-side networks located at individual sites is also being activated to overcome regional limitations of source-side detection. In this paper, we evaluate the performance of a collaborative source-side DDoS attack detection using statistical weights. The statistical weight is calculated based on the detection rate and false positive rate corresponding to the time zone of the individual source-side network. By calculating weighted sum of the source-side DoS attack detection results from various sites, the proposed method determines whether a DDoS attack happens. As a result of the experiment based on actual DNS request to traffic, it was confirmed that the proposed technique reduces false positive rate 2% while maintaining a high attack detection rate.