Vol. 47, No. 1,
Jan. 2020
Digital Library
A Study on Parallel Optimization of Regional Ocean Model Using KISTI 5th Supercomputer Nurion System
Dong-Hoon Kim, Chaewook Lim, Min-Su Joh, Jooneun An, Il-Ju Moon, Seung-Buhm Woo
http://doi.org/10.5626/JOK.2020.47.1.1
A parrallel optimiztion of the Regional Ocean Modeling system (ROMS) was conducted by analyzing the computational characteristics of the Knights Landing (KNL) system of Nurion, which is the fifth supercomputer at the Korea Institute of Science and Technology Information (KISTI). Although the KNL system comprises more than twice as many cores per node as the Skylake system, the performance of the KNL system is known to be about three times slower than that of Skylake system, when using the same number of cores. However, the KNL performance optimized for the ROMS model in Nurion is only about twice that of the Skylake system suggesting that the KNL system of Nurion is approximately 1.3-fold faster than the normal KNL system. In this study, two types of numerical experiments (ideal & real cases) were conducted to compare the computational capabilities of the KNL and SKL systems; The KNL system has more computational efficiency. The KNL system shows continuous improvement in performance even in maximum core experiments under both ideal and real case simulation, which is an advantage for the simulation of super parallel numerical calculation.
Deadline Task Scheduling for Mitigating the CPU Performance Interference in Android Systems
Jeongwoong Lee, Taehyung Lee, Young Ik Eom
http://doi.org/10.5626/JOK.2020.47.1.11
In the Android Linux kernel, most of the tasks are expected to run fairly, and so, there can be delays in processing time-sensitive applications. In particular, since the user may feel inconveniences when the delay occurs in media data processing or biometrics processing such as fingerprint recognition, the tasks requiring completion within a given time should be considered as deadline tasks. However, using the deadline scheduler in current Android systems can cause two problems. First, as deadline tasks come to the system and are executed, the CPU energy consumption can be increased. Second, the high priority of the deadline tasks can cause performance degradation of the normal tasks. To mitigate these problems, this paper proposes a method of scheduling deadline tasks on Android systems, which reduces the performance impact on normal tasks, while trying to minimize energy consumption. Our evaluation on the CPU benchmark shows that the proposed method improves the CPU performance by about 10% compared with the conventional deadline scheduler, but does not increase power consumption by effectively utilizing CPU frequency.
CAAM - Model for National-level Cyber Attack Attribution
Min-ho Lee, Chang-wook Park, Wan-ju Kim, Jae-sung Lim
http://doi.org/10.5626/JOK.2020.47.1.19
Recently, security companies have been reporting that some organizations engaging in carry out cyber attacks are suspected of receiving state-sponsored support. To effectively respond to these cyber-attack groups, it is critical to detect and quickly analyze the characteristics of the attacks to identify the countries responsible first for such terroristic acts. This paper presents the attribution model (CAAM) for state-sponsored cyber attacks, and CAAM analyzes the characteristics of such cyber attacks through the four-step process of detection and collection, analysis, evaluation and visualization. The detailed elements for analyzing the characteristics of cyber attacks were divided into five categories: Tools and technology of attack organizations, Infrastructure of attack organizations, Structure of malicious codes, Motivation of attacks, and External factors. Five factors were assessed by country to identify those that support cyber attacks. The application of CAAM is expected to enable rapid analysis of state-sponsored cyber attacks and has validated the effectiveness of the CAAM model through comparison with the existing attack group analysis model.
Exploiting Inverse Power of Two Non-Uniform Quantization Method to Increase Energy Efficiency in Deep Neural Networks
http://doi.org/10.5626/JOK.2020.47.1.27
DNN"s computational complexity makes it difficult for application to embedded devices of limited resources because the deep learning requires high performance computing power and consumes considerable energy. To mitigate this, this paper proposes an energy-efficient Inverse Power of Two(IPow2) nonuniform quantization technique to induce more sparsity than the existing quantization methods while reducing precision of weights, resulting in the reduction of computational complexity as well as energy consumption in DNN. Accuracy and energy efficiency of the proposed IPow2 are quantitatively validated by executing image classification task with data sets of CIFAR- 10/ImageNet through implementing the quantized AlexNet/VGGNet models of a variety of mapping policies. Experimental results show that the proposed IPow2 method consumes less energy by 63.2% and 66.5% while achieving minor accuracy loss by 2.2% and 2.5% respectively compared with the full precision one, in case of two-bits quantization in the AlexNet/VGGNet models.
Automatic Text Summarization Based on Selective OOV Copy Mechanism with BERT Embedding
http://doi.org/10.5626/JOK.2020.47.1.36
Automatic text summarization is a process of shortening a text document via extraction or abstraction. Abstractive text summarization involves using pre-generated word embedding information. Low-frequency but salient words such as terminologies are seldom included in dictionaries, that are so called, out-of-vocabulary (OOV) problems. OOV deteriorates the performance of the encoder-decoder model in the neural network. To address OOV words in abstractive text summarization, we propose a copy mechanism to facilitate copying new words in the target document and generating summary sentences. Different from previous studies, the proposed approach combines accurately pointing information, selective copy mechanism, embedded by BERT, randomly masking OOV, and converting sentences from morpheme. Additionally, the neural network gate model to estimate the generation probability and the loss function to optimize the entire abstraction model was applied. Experimental results demonstrate that ROUGE-1 (based on word recall) and ROUGE-L (longest used common subsequence) of the proposed encoding-decoding model have been improved at 54.97 and 39.23, respectively.
Latent Dirichlet Allocation Based Crime Code Clustering and Crime Prediction
http://doi.org/10.5626/JOK.2020.47.1.45
Predicting crime using crime data has become one of the most actively researched disciplines in major cites. Based on the research, law enforcement officials are shifting their efforts from crime investigation to crime prevention through predictive policing. Predictive policing highly relies on mathematics and statistics and identifies the underlying patterns of crimes. Based on these patterns, law enforcement officials can identify potential criminal activities. For accurate prediction, crime data must be well organized and managed. We first introduce one of the popular crime data set actively used by researchers. The data set categorizes each incident through a crime code. Examining the frequency of these codes allows regional agencies to predict the type of potential crimes, leading to effective predictive policing. In this research, we introduce a machine learning-based approach that can identify the similarity between the codes. Based on these similarities, we compute the frequencies of clusters and predict the code of potential crimes. Our experimental results show how our algorithm outperforms the statistical method.
Parser Generators Sharing LR Automaton Generators and Accepting General Purpose Programming Language-based Specifications
Jintaeck Lim, Gayoung Kim, Seunghyun Shin, Kwanghoon Choi, Iksoon Kim
http://doi.org/10.5626/JOK.2020.47.1.52
This paper proposes two ways to develop LR parsers easily. First, one can write a parser specification in a general programming language and derive the benefits of syntax error checking, code completion, and type-error checking over the specification from the language’s development environment. Second, to make it easy to develop a parser tool for a new programming language, the automata generation for the parser specifications is in a modular form. With the idea proposed in this study, we developed a tool for writing parsers in Python, Java, C++, and Haskell. We also demonstrated the two aforementioned advantages in an experiment.
Recommending Similar Users Through Interaction Analysis in Social IoT Environments
Yeondong Kim, Dojin Choi, Jongtae Lim, Kyoungsoo Bok, Jaesoo Yoo
http://doi.org/10.5626/JOK.2020.47.1.61
Recently, there has been extensive research on the social internet of things(Social IoT) that combines social networks and internet of things. Social IoT is integral for the connection between as well as for establishing relationships between users and objects for sharing information between objects or users. In this paper, we propose a method that recommends similar users by considering interaction between objects and users in the social IoT environments. The similar users can be found by analyzing the behavior of the users around the object. The proposed method improves the accuracy of similarity by calculating similarity in determining interests based on documents written by users in social networks. Finally, it recommends Top-N users as similar users based on the two similarity values. To show the superiority of the proposed method, we conducted various performance evaluations.
Korean Morphological Analyzer for Neologism and Spacing Error based on Sequence-to-Sequence
Byeongseo Choe, Ig-hoon Lee, Sang-goo Lee
http://doi.org/10.5626/JOK.2020.47.1.70
In order to analyze Internet text data from Korean internet communities, it is necessary to accurately perform morphological analysis even in a sentence with a spacing error and adequate restoration of original form for an out-of-vocabulary input. However, the existing Korean morphological analyzer often uses dictionaries and complicate preprocessing for the restoration. In this paper, we propose a Korean morphological analyzer model which is based on the sequence-to-sequence model. The model can effectively handle the spacing problem and OOV problem. In addition, the model uses syllable bigram and grapheme as additional input features. The proposed model does not use a dictionary and minimizes rule-based preprocessing. The proposed model showed better performance than other morphological analyzers without a dictionary in the experiment for Sejong corpus. Also, better performance was evident for the dataset without space and sample dataset collected from Internet.
Analysis of Reward Functions in Deep Reinforcement Learning for Continuous State Space Control
http://doi.org/10.5626/JOK.2020.47.1.78
Deep Reinforcement Learning (DRL), which uses deep neural networks for the approximation of the value function and the policy, in continuous state-space control tasks has recently shown promising results. However, the use of deep neural networks as function approximators has often resulted in intractable analyses of DRL algorithms mainly due to their non-convexities and thus a lack of theoretical guarantee such as asymptotic global convergence of the learning algorithm. Considering the fact that the reward function in reinforcement learning is one of the key entities that determines the overall characteristics of the learning agents, we focused on a smaller but an important aspect of the analysis, investigating the structure of widely used reward functions in DRL tasks and their possible effects on the learning algorithm. The proposed analysis may facilitate identification of appropriate reward functions in DRL tasks, which has often been conducted via trial and error.
Cross-Entropy Planning with Prior Updates
HyeongJoo Hwang, Youngsoo Jang, Jaeyoung Park, Kee-Eung Kim
http://doi.org/10.5626/JOK.2020.47.1.88
This paper introduces a method of cross-entropy planning which updates prior probability for planning optimization process. Cross-entropy planning is a popular method in online planning and involves the extraction of samples from a simulation environment and selection of optimal action based on the values of the extracted samples. The performance of the cross-entropy planning is limited due to involvement of optimization processes without usage of previous planning results. We propose a method that updates prior probabilities for the optimization process based on the action sequences acquired from the cross-entropy planning. The proposed method improves the performance of cross-entropy planning with progression of planning epoch. We evaluated the proposed method based on the comparison with the cross-entropy planning in a physical-based simulation (OpenAI Gym) environment.
An Efficient Distributed In-memory High-dimensional Indexing Scheme for Content-based Image Retrieval in Spark Environments
Dojin Choi, Songhee Park, Yeondong Kim, Jiwon Wee, Hyeonbyeong Lee, Jongtae Lim, Kyoungsoo Bok, Jaesoo Yoo
http://doi.org/10.5626/JOK.2020.47.1.95
Content-based image retrieval that searches an object in images has been utilizing for criminal activity monitoring and object tracking in video. In this paper, we propose a high-dimensional indexing scheme based on distributed in-memory for the content-based image retrieval. It provides similarity search by using massive feature vectors extracted from images or objects. In order to process a large amount of data, we utilized a big data platform called Spark. Moreover, we employed a master/slave model for efficient distributed query processing allocation. The master distributes data and queries. and the slaves index and process them. To solve k-NN query processing performance problems in the existing distributed high-dimension indexing schemes, we propose optimization methods for the k-NN query processing considering density and search costs. We conduct various performance evaluations to demonstrate the superiority of the proposed scheme.
Enhancing Throughput of Sensor Data Acquisition using Mobile Device Information in Internet of Things
Cheonyong Kim, Sangdae Kim, Seungmin Oh, Kwansoo Jung
http://doi.org/10.5626/JOK.2020.47.1.109
Internet of Things (IoT) is a technology to provide emerging applications, such as remote data collection. Recently, the active deployment of wireless sensors and the explosive increase in the number of mobile devices has enabled opportunistic data collection (ODC), in which mobile devices collect data from distributed sensors. In ODC, the sensor data can be collected without the additional overhead of installing and operating a network infrastructure. However, the time duration for data transfer cannot be guaranteed due to the unpredictable mobility of mobile devices in ODC. Therefore, existing studies have used a conservative flow control mechanism focusing on reliable data transfer. However, the conservative flow control degrades the data collection performance when the time duration for data transfer is sufficiently long. In this paper, an efficient flow control scheme was proposed for enhancing the data collection performance of ODC. The proposed scheme is based on the communication time between a mobile device and a sensor. The efficiency-oriented flow control mechanism is applied with a long communication time, while a short communication time leads to the reliability-oriented flow control mechanism. By using the proposed scheme, the data collection performance can be enhanced without degrading the reliability of data transfer.
Search
Journal of KIISE
- ISSN : 2383-630X(Print)
- ISSN : 2383-6296(Electronic)
- KCI Accredited Journal
Editorial Office
- Tel. +82-2-588-9240
- Fax. +82-2-521-1352
- E-mail. chwoo@kiise.or.kr