Vol. 46, No. 3,
Mar. 2019
Digital Library
Performance and Energy Comparison of Different BLAS and Neural Network Libraries for Efficient Deep Learning Inference on ARM-based IoT Devices
http://doi.org/10.5626/JOK.2019.46.3.219
Cloud computing is generally used to perform deep learning on IoT devices. However, its application is associated with limitations such as connection instability, energy consumption for communication, and security vulnerabilities. To solve such problems, recent attempts at performing deep learning within IoT devices have occurred. These attempts mainly suggest either lightweight deep learning models or compression techniques concerning IoT devices, but they lack analysis of the effect when it is performed in actual IoT devices. Since each IoT device has different configuration of processing units and supported libraries, it is necessary to analyze various execution environments in each IoT device in order to perform optimized deep learning. In this study, performance and energy of IoT devices with various hardware configurations were measured and analyzed according to the application of the deep learning model, library, and compression technique. It was established that utilizing the appropriate libraries improve both speed and energy efficiency up to 13.3 times and 48.5 times, respectively.
Design of a ModelAgent for Efficient Development of Dynamic Models
Inhan Kim, Younghwan Jeong, Won-sik Lee, Soung-Hyouk Wi, Hamin Jeong
http://doi.org/10.5626/JOK.2019.46.3.228
Dynamic models are important elements in the process of developing weapon system simulators based on the Defense Modeling and Simulation(DM&S). Also, the model’s accuracy according to the purpose of the DM&S is a very important factor, therefore the modules of the dynamic models should be modified to satisfy required model’s fidelity. The objective of this study is to design a structure of ModelAgent that can efficiently develop and manage the dynamic models. The ModelAgent is designed not only to improve reuse by encapsulating modules using object oriented language but also to be scalable to accommodate a variety of requirements by applying design patterns. Also, the ModelAgent is designed to provide a common interface that enhances portability in the development of dynamic models.
Korean Machine Reading Comprehension using S³-Net based on Position Encoding
Choeneum Park, Changki Lee, Hyunki Kim
http://doi.org/10.5626/JOK.2019.46.3.234
S³-Net is a deep learning model that is used in machine reading comprehension question answering (MRQA) based on Simple Recurrent Unit and Self-Matching Networks that calculates attention weight for own RNN sequence. The answers to the questions in the MRQA occur within the passage, because any passage is made up of several sentences, so the length of the input sequence becomes longer and the performance deteriorates. In this paper, a hierarchical model that adds sentence-level encoding and S³-Net that applies position encoding to check word order information to solve the problem of long-term context degradation are proposed. The experimental results show that the S³-Net model proposed in this paper has a performance of 69.43% in EM and 81.53% in F1 for single test, and 71.28% in EM and 82.67 in F1 for ensemble test.
Sentence Similarity Prediction based on Siamese CNN-Bidirectional LSTM with Self-attention
Mintae Kim, Yeongtaek Oh, Wooju Kim
http://doi.org/10.5626/JOK.2019.46.3.241
A deep learning model for semantic similarity between sentences was presented. In general, most of the models for measuring similarity word use level or morpheme level embedding. However, the attempt to apply either word use or morpheme level embedding results in higher complexity of the model due to the large size of the dictionary. To solve this problem, a Siamese CNN-Bidirectional LSTM model that utilizes phonemes instead of words or morphemes and combines long short term memory (LSTM) with 1D convolution neural networks with various window lengths that bind phonemes is proposed. For evaluation, we compared our model with Manhattan LSTM (MaLSTM) which shows good performance in measuring similarity between similar questions in the Naver Q&A dataset (similar to Kaggle Quora Question Pair).
Effective Generative Chatbot Model Trainable with a Small Dialogue Corpus
Jintae Kim, Hyeon-gu Lee, Harksoo Kim
http://doi.org/10.5626/JOK.2019.46.3.246
Contrary to popular retrieval-based chatbot models, generative chatbot models do not depend on predefined responses, but rather generate new responses based on well-trained neural networks. However, they require a large number of training corpus in the form of query-response pairs. If the training corpus are insufficient, they make grammatical errors emanating from out-of-vocabulary or sparse data problems, mostly in longer sentences. To overcome this challenge, we proposed a chatbot model based on sequence-to-sequence neural network using a mixture of words and syllables as encoding-decoding units. Moreover, we proposed a two-step training procedure involving pre-training using a large non-dialogue corpus and retraining using a smaller dialogue corpus. In the experiment involving small dialogue corpus (47,089 query-response pairs for training and 3,000 query-response pairs for evaluation), the proposed encoding-decoding units resulted to a reduction in out-of-vocabulary problem while the two-step training method led to improved performance measures like BLEU and ROUGE.
C++ based General-purpose Open Source Deep Learning Framework, WICWIU
Chunmyong Park, Jeewoong Kim, Yunho Kee, Jihyeon Kim, Seonggyeol Yoon, Eunseo Choi, Injung Kim
http://doi.org/10.5626/JOK.2019.46.3.253
In this paper, we introduce WICWIU, the first open source deep learning framework among Korean universities. WICWIU provides a variety of operators and modules together with a network structure that can represent an arbitrary general computational graph. The WICWIU features are sufficient to compose widely used deep learning models such as Inception, ResNet, and DenseNet. WICWIU also supports GPU-based massive parallel computing which significantly accelerates the training of neural networks. It is also easily accessible for C++ developers because the whole API is provided in C++. WICWIU has an advantage over Python-based frameworks in memory and performance optimization based on the C++ environment. This eases the customizability of WICWIU for environments with limited resources. WICWIU is readable and extensible because it is composed of C++ codes coupled with consistent APIs. With Korean documentation, it is particularly suitable for Korean developers. WICWIU applies the Apache 2.0 license which is available for any research or commercial purposes for free.
Implementation of Web-based High-Throughput Screening Calculation System
Seong-ho Cho, Inhee Kim, Jiwon Kong, Daesan Kim, Namhoon Kwon, Robert E. Burrier, Sunghoon Kim
http://doi.org/10.5626/JOK.2019.46.3.260
In this paper, we describe the development of a web-based HTSCS (High Throughput Screening Calculation System), which is the process for assisting high-speed screening and calculation of drug candidates based on biochemical reactions from Nanoluciferase system. This system calculates and displays large amounts of experimental data with single-dose and multi-dose concentration in-vitro cell experiments through charts and graphs. Therefore, we can utilize the HTSCS to save time and effort required for drug screening.
Quality Adaptation Scheme Based on Stability to Improve QoE of HTTP Adaptive Streaming in Wireless Networks
Minsu Kim, Heekwang Kim, Kwangsue Chung
http://doi.org/10.5626/JOK.2019.46.3.268
With the recent development of networks, the demand for video streaming has also been increasing, and more attention has been focused on HTTP adaptive streaming. HTTP adaptive streaming can guarantee QoE (Quality of Experience) because the client is provided with the options of selecting the quality adaptively according to the network state. However, in wireless networks, bandwidth is inaccurately measured due to high delay and packet loss. The existing quality adaptation schemes degrade the QoE because these schemes select quality using the measured bandwidth. In this paper, a quality adaptation scheme based on stability is proposed to improve QoE of HTTP adaptive streaming. The proposed scheme calculates the buffer underflow probability and the instability using the changes in the video quality and buffer and selects the quality according to the defined quality adaptation region. Experimental results established that the proposed scheme improves QoE as compared to the existing schemes by high average video quality and fewer quality changes.
RSU-independent Message Authentication Scheme using CRT-based Group Key in VANET
Jin Sook Bong, Yu Hwa Suh, Ui Jin Jang, Yongtae Shin
http://doi.org/10.5626/JOK.2019.46.3.277
Security in communication is still an important issue because information transmitted via VANET provides safety-related services to drivers. In this context, there have been many studies related to the sending and receiving messages securely in VANET. In many studies, the RSU was assumed as a trust or semi-trust device and RSU played an important role in message authentication. However, a large number of RSUs cannot be considered trustworthy because the security of the RSU cannot be secured. Therefore, this paper proposes a message authentication scheme that is not dependent on the RSU. For this, a group key based on CRT (Chinese Remainder Theorem) is used. In the proposed scheme, the CRT-based key server calculates and distributes a `private key` and a `group key seed` to each vehicle. Then, each vehicle extracts the group key from the group key seed and uses it to authenticate the message. The proposed scheme eliminates the RSU dependence in a message verification. And it is also possible to easily withdraw a vehicle from a group.
Document Summarization Using TextRank Based on Sentence Embedding
Seok-won Jeong, Jintae Kim, Harksoo Kim
http://doi.org/10.5626/JOK.2019.46.3.285
Document summarization is creating a short version document that maintains the main content of original document. An extractive summarization has been actively studied by the reason of it guarantees the basic level of grammar and high level of accuracy by copying a large amount of text from the original document. It is difficult to consider the meaning of sentences because the TextRank, which is a typical extractive summarization method, calculates an edge of graph through the frequency of words. In a bid to solve these drawbacks, we propose a new TextRank using sentence embedding. Through experiments, we confirmed that the proposed method can consider the meaning of the sentence better than the existing method.
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