• Journal of KIISE
• /
• v.42 no.6
• /
• pp.791-800
• /
• 2015

In this paper, we propose the contribution-level-based opportunistic flooding in a wireless multihop network which achieves outstanding transmission efficiency and reliability. While the potential of the the predetermined relay node to fail in its receipt of broadcast packets is due to the inherent instability in wireless networks, our proposed flooding actually increases network reliability by applying the concept of opportunistic routing, whereby relay-node selection is dependent on the transmission result. Additionally, depending on the contribution level for the entire network, the proposed technique enhances transmission efficiency through priority adjustment and the removal of needless relay nodes. We use the NS-3 simulator to compare the proposed scheme with dominant pruning. The analysis results show the improved performance in both cases: by 35% compared with blind flooding from the perspective of the transmission efficiency, and by 20~70% compared to dominant pruning from the perspective of the reliability.

• Journal of Korea Multimedia Society
• /
• v.4 no.6
• /
• pp.524-533
• /
• 2001

Efficient continuous speech recognition system for practical applications requires that the processing be carried out in real time and high recognition accuracy. In this paper, we study the acoustic models by adopting the PDT-SSS algorithm and the language models by iterative learning so as to improve the speech recognition accuracy. And the adaptive pruning algorithm is applied to the continuous speech. To verify the effectiveness of proposed method, we carried out the continuous speech recognition for the Korean air flight reservation task. Experimental results show that the adopted algorithm has the average 90.9% for continuous speech recognition and the average 90.7% for word recognition accuracy including continuous speech. And in case of adopting the adaptive pruning algorithm to continuous speech, it reduces the recognition time of about 1.2 seconds(15%) without any loss of accuracy. From the result, we proved the effectiveness of the PDT-SSS algorithm and the adaptive pruning algorithm.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2016.04a
• /
• pp.55-58
• /
• 2016

스카이라인 질의란 다수의 선택지 중 '선호될 만한(preferable)' 선택지를 요청하는 질의이다. 사용자가 검토해야하는 선택지의 수를 대폭 감소시키는 스카이라인 질의는 데이터가 폭증하는 빅데이터 환경에서 매우 유용하게 활용된다. 이러한 배경에서 대용량 데이터에 대한 스카이라인 질의를 분산 병렬 처리하는 기법이 각광을 받고 있으며, 특히 맵리듀스(MapReduce) 기반의 분산 병렬 처리 기법 연구가 활발히 진행 중이다. 맵리듀스 기반 알고리즘의 병렬성 제고를 위해서는 부하 불균등 문제 중복 계산 문제 과다한 네트워크 비용 발생 문제를 해소해야 한다. 최근 각 기반 공간분할 기법을 사용하여 부하 불균등 문제와 중복 계산 문제를 해소하는 맵리듀스 기반 스카이라인 질의 처리 기법이 제안되었으나 해당 기법은 네트워크 비용 관점에서 최적화되어있지 않다. 본 논문에서는 부하 불균등 문제와 중복 계산 문제를 해소하면서도 프루닝을 통해 네트워크 비용 절감 시킬 수 있는 새로운 맵리듀스 기반 병렬 스카이라인 질의 처리 기법인 MR-SEAP(MapReduce sample Skyline object Equality Angular Partitioning)을 제안한다. MR-SEAP에서는 데이터를 샘플링하여 샘플 스카이라인 객체를 추출한 뒤 해당 객체들을 균등 분배하는 각도를 기준으로 공간을 분할하여 스카이라인 질의를 병렬 계산하되, 샘플 스카이라인을 이용하여 다수의 객체를 사전에 프루닝함으로써 네트워크 비용을 절감한다. 본 논문에서는 다양한 데이터 수량(cardinality) 및 분포(distribution)에 따른 제안 기법의 성능을 실험 평가함으로써 제안 기법의 우수성을 검증한다.

• KIPS Transactions on Software and Data Engineering
• /
• v.6 no.1
• /
• pp.1-8
• /
• 2017

Given a multi-dimensional dataset of tuples, a skyline query returns a subset of tuples which are not 'dominated' by any other tuples. Skyline query is very useful in Big data analysis since it filters out uninteresting items. Much interest was devoted to the MapReduce-based parallel processing of skyline queries in large-scale distributed environment. There are three requirements to improve parallelism in MapReduced-based algorithms: (1) workload should be well balanced (2) avoid redundant computations (3) Optimize network communication cost. In this paper, we introduce MR-SEAP (MapReduce sample Skyline object Equality Angular Partitioning), an efficient angular space partitioning based skyline query processing using sampling-based pruning, which satisfies requirements above. We conduct an extensive experiment to evaluate MR-SEAP.

• Journal of the Institute of Convergence Signal Processing
• /
• v.22 no.3
• /
• pp.99-103
• /
• 2021

Recently, convolutional neural networks (CNNs) have been widely utilized due to their outstanding performance in various computer vision fields. However, due to their computational-intensive and high memory requirements, it is difficult to deploy CNNs on hardware platforms that have limited resources, such as mobile devices and IoT devices. To address these limitations, a neural network compression research is underway to reduce the size of neural networks while maintaining their performance. This paper proposes a CNN compression technique that dynamically adjusts the thresholds of pruning, one of the neural network compression techniques. Unlike the conventional pruning that experimentally or heuristically sets the thresholds that determine the weights to be pruned, the proposed technique can dynamically find the optimal thresholds that prevent accuracy degradation and output the light-weight neural network in less time. To validate the performance of the proposed technique, the LeNet was trained using the MNIST dataset and the light-weight LeNet could be automatically obtained 1.3 to 3 times faster without loss of accuracy.

• The Journal of the Acoustical Society of Korea
• /
• v.22 no.6
• /
• pp.457-462
• /
• 2003

In this paper, we propose a speaker pruning method for real-time processing and improving performance of speaker identification system based on GMM(Gaussian Mixture Model). Conventional speaker identification methods, such as ML (Maximum Likelihood), WMR(weighting Model Rank), and MWMR(Modified WMR) we that frame likelihoods are calculated using the whole frames of each input speech and all of the speaker models and then a speaker having the biggest accumulated likelihood is selected. However, in these methods, calculation cost and processing time become larger as the increase of the number of input frames and speakers. To solve this problem in the proposed method, only a part of speaker models that have higher likelihood are selected using only a part of input frames, and identified speaker is decided from evaluating the selected speaker models. In this method, fm can be applied for improving the identification performance in speaker identification even the number of speakers is changed. In several experiments, the proposed method showed a reduction of 65% on calculation cost and an increase of 2% on identification rate than conventional methods. These results means that the proposed method can be applied effectively for a real-time processing and for improvement of performance in speaker identification.

• Journal of KIISE:Software and Applications
• /
• v.30 no.3_4
• /
• pp.326-338
• /
• 2003

The objective of a neural network design and model selection is to construct an optimal network with a good generalization performance. However, training data include noises, and the number of training data is not sufficient, which results in the difference between the true probability distribution and the empirical one. The difference makes the teaming parameters to over-fit only to training data and to deviate from the true distribution of data, which is called the overfitting phenomenon. The overfilled neural network shows good approximations for the training data, but gives bad predictions to untrained new data. As the complexity of the neural network increases, this overfitting phenomenon also becomes more severe. In this paper, by taking statistical viewpoint, we proposed an integrative process for neural network design and model selection method in order to improve generalization performance. At first, by using the natural gradient learning with adaptive regularization, we try to obtain optimal parameters that are not overfilled to training data with fast convergence. By adopting the natural pruning to the obtained optimal parameters, we generate several candidates of network model with different sizes. Finally, we select an optimal model among candidate models based on the Bayesian Information Criteria. Through the computer simulation on benchmark problems, we confirm the generalization and structure optimization performance of the proposed integrative process of teaming and model selection.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.6
• /
• pp.850-858
• /
• 2022

To deploy Gate Recurrent Units (GRU) on resource-constrained embedded devices, this paper presents a reconfigurable FPGA-based GRU accelerator that enables structured compression. Firstly, a dense GRU model is significantly reduced in size by hybrid quantization and structured top-k pruning. Secondly, the energy consumption on external memory access is greatly reduced by the proposed reuse computing pattern. Finally, the accelerator can handle a structured sparse model that benefits from the algorithm-hardware co-design workflows. Moreover, inference tasks can be flexibly performed using all functional dimensions, sequence length, and number of layers. Implemented on the Intel DE1-SoC FPGA, the proposed accelerator achieves 45.01 GOPs in a structured sparse GRU network without batching. Compared to the implementation of CPU and GPU, low-cost FPGA accelerator achieves 57 and 30x improvements in latency, 300 and 23.44x improvements in energy efficiency, respectively. Thus, the proposed accelerator is utilized as an early study of real-time embedded applications, demonstrating the potential for further development in the future.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2018.07a
• /
• pp.427-430
• /
• 2018

본 논문에서는 오목을 두는 지능적 프로그램을 설계하고 구현하였다. 규칙은 렌주 룰(renju rule)을 기준으로 하였으며, $15{\times}15$ 게임 판에서 오목을 둔다. 초기에는 문제 분석을 통하여 분석된 가중치로 판단을 하여 판단을 하여 게임을 진행하도록 하였으나, 반복된 수행의 경험적 판단을 통하여 얻은 정보로 여러 차례 수정하며 고정된 가중치를 구성하고, 이 가중치를 게임에서 돌을 놓을 때 평가 기준으로 삼도록 하였으며, 최소-최대 게임 트리(min-max game tree)를 이용하여 상대가 있는 게임을 수행할 수 있도록 하였다. 또한 프로그램 자신에게 유리한 수를 찾기 위한 탐색에서 무의미한 노드들의 전개를 줄여 제한된 시간안에 좋은 수를 찾을 수 있도록 알파 베타 가지치기(alpha-beta pruning)를 사용하도록 프로그램을 구현하였다. 이렇게 구현된 오목 프로그램은 게임을 본 프로그램과 게임 하기 원하는 주변의 일반인들에게 90% 이상의 승률을 보이고 있었다.

• The Journal of Korea Robotics Society
• /
• v.18 no.1
• /
• pp.88-92
• /
• 2023

Lithium-ion batteries are actively used in various industrial sites such as field robots, drones, and electric vehicles due to their high energy efficiency, light weight, long life span, and low self-discharge rate. When using a lithium-ion battery in a field, it is important to accurately estimate the SoC (State of Charge) of batteries to prevent damage. In recent years, SoC estimation using data-based artificial neural networks has been in the spotlight, but it has been difficult to deploy in the embedded board environment at the actual site because the computation is heavy and complex. To solve this problem, neural network lightening technologies such as network pruning have recently attracted attention. When pruning a neural network, the performance varies depending on which layer and how much pruning is performed. In this paper, we introduce an optimized pruning technique by improving the existing pruning method, and perform a comparative experiment to analyze the results.