• Journal of the Institute of Electronics and Information Engineers
• /
• v.50 no.1
• /
• pp.166-173
• /
• 2013

As the number of IPs and the communication volume among them have constantly increased, on-chip crossbar network is now the most widely-used on-chip communication backbone of contemporary SoCs. The on-chip crossbar network consists of multiple crossbars and the connections among the IPs and the crossbars. As the complexity of SoCs increases, it has also become more and more complex to determine the topology of the crossbar network. To tackle this problem, this paper proposes an on-chip crossbar network topology method for application-specific systems. The proposed method uses mixed integer linear programming to solve the topology synthesis problem, thus the global optimality is guaranteed. Unlike the previous MILP-based methods which represent the topology with adjacency matrixes of IPs and crossbar switches, the proposed method uses the communication edges among IPs as the basic element of the representation. The experimental results show that the proposed MILP formulation outperforms the previous one by improving the synthesis speed by 77.1 times on average, for 4 realistic benchmarks.

• KNOM Review
• /
• v.24 no.2
• /
• pp.24-34
• /
• 2021

Service providers using edge computing provide a high level of service. As a result, devices store important information in inner storage and have become a target of the latest cyberattacks, which are more difficult to detect. Although experts use a security system such as intrusion detection systems, the existing intrusion systems have low detection accuracy. Therefore, in this paper, we proposed a machine learning model for more accurate intrusion detections of devices in edge computing. The proposed model is a hybrid model that combines a stacked sparse autoencoder (SSAE) and a convolutional neural network (CNN) to extract important feature vectors from the input data using sparsity constraints. To find the optimal model, we compared and analyzed the performance as adjusting the sparsity coefficient of SSAE. As a result, the model showed the highest accuracy as a 96.9% using the sparsity constraints. Therefore, the model showed the highest performance when model trains only important features.

• Transactions on Semiconductor Engineering
• /
• v.2 no.1
• /
• pp.17-20
• /
• 2024

This paper presents a novel lightweight object detection model tailored for low-powered edge devices, addressing the limitations of traditional resource-intensive computer vision models. Our proposed detector, inspired by the Single Shot Detector (SSD), employs a compact yet robust network design. Crucially, it integrates an 'enhancer block' that significantly boosts its efficiency in detecting smaller objects. The model comprises two primary components: the Light_Block for efficient feature extraction using Depth-wise and Pointwise Convolution layers, and the Enhancer_Block for enhanced detection of tiny objects. Trained from scratch on the Udacity Annotated Dataset with image dimensions of 300x480, our model eschews the need for pre-trained classification weights. Weighing only 5.5MB with approximately 0.43M parameters, our detector achieved a mean average precision (mAP) of 27.7% and processed at 140 FPS, outperforming conventional models in both precision and efficiency. This research underscores the potential of lightweight designs in advancing object detection for edge devices without compromising accuracy.

• Annual Conference of KIPS
• /
• 2017.04a
• /
• pp.1224-1227
• /
• 2017

포그 컴퓨팅 기술은 물리적 환경과 빈번하게 상호작용이 일어나는 사이버-물리 시스템에서 네트워크의 엣지에 있는 시스템이 컴퓨팅 작업을 수행하도록 함으로써 지역의 데이터를 실시간으로 수집하고 처리할 수 있다. 본 논문에서는 스쿨존내에서 안전을 높이기 위한 횡단보도의 신호등에 포그 컴퓨팅 기술을 적용한다. 신호등 시스템은 횡단보도에 접근하는 자동차를 인지하고, 위험 상황을 미리 방지하기 위해 자동차를 제어할 수 있다. 실험을 위해 사물인터넷 기술을 이용해 소형 테스트베드를 만들었으며, 신호 정보를 변화시키며 실험을 수행한다.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2016.01a
• /
• pp.57-58
• /
• 2016

ICT의 보급, 확대는 데이터 센터의 중요성을 높이고 보다 성능이 좋으며 체적 당 소비전력이 큰 서버를 수용할 수 있는 데이터 센터의 수요를 창출하고 있다. 현재 데이터 센터는 데이터 센터 활용 시에 구성요소들에 대한 상당한 대역폭을 필요로 하나 현 데이터센터에 적용된 토폴로지는 고성능 IP 스위치/라우터를 사용하더라도 네트워크 엣지 계층에서는 기본 활용도의 50%의 bandwidth밖에 지원하지 못한다. 따라서 이러한 문제를 해결하기 위해 OMNeT++을 이용하여 데이터 센터 토폴로지 중 하나인 Fat-tree를 모델링하고 데이터 센터 제반 환경을 구축, latency, power consumption, heat dissipation 등의 기준지표를 성능평가 하였다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2022.06a
• /
• pp.1322-1324
• /
• 2022

최근 딥러닝의 발전에 따라 단일 이미지 초해상도 분야에 좋은 성과를 보여주고 있다. 그러나 보다 더 높은 성능을 획득하기 위해 네트워크의 깊이 및 파라미터의 수가 크게 증가하였고, 모바일 및 엣지 디바이스에 원활하게 적용되기 위하여 딥러닝 모델 경량화의 필요성이 대두되고 있다. 이에 본 논문에서는 초해상도 모델 중 하나인 EDSR(Enhanced Deep Residual Network)에 대조 학습 기반 지식 전이를 적용한 경량화 기법을 제안한다. 실험 결과 제안한 지식 전이 기법이 기존의 다른 지식 증류 기법보다 향상된 성능을 보임을 확인하였다.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2021.10a
• /
• pp.313-316
• /
• 2021

In the IoT(IoT: Internet of Things) environment, network configuration is essential to collect data generated from objects. Various communication methods are used to process data of objects, and wireless communication methods such as Bluetooth and WiFi are mainly used. In order to collect data of objects, a communication module must be installed to collect data generated from sensors or edge devices in real time. And in order to deliver data to the database, a software architecture must be configured. Data generated from objects can be stored and managed in a database in real time, and data necessary for industrial safety can be extracted and utilized for industrial safety service applications. In this paper, a network environment was constructed using a LoRa(LoRa: Long Range) gateway to collect object data, and a client/server data collection model was designed to collect object data transmitted from the LoRa module. In order to secure the resources necessary for data collection and storage management without data leakage, data collection should be possible in real time. As an application service, location data required for industrial safety can be stored and managed in a database in real time.

• Journal of IKEEE
• /
• v.28 no.3
• /
• pp.458-464
• /
• 2024

This study proposes a fault-tolerant caching system to address the issue of caching content imbalance caused by the dynamic departure and participation of cache servers in a heterogeneous cache server network, and validates it in both real and virtual environments. With the increase of large-scale media content requiring various types and resolutions, the necessity of cache servers as key components to reduce response time to user requests and alleviate network load has been growing. In particular, research on heterogeneous cache server networks utilizing edge computing and low-power devices has been actively conducted recently. However, in such environments, the irregular departure and participation of cache servers can occur frequently, leading to content imbalance among the cache servers deployed in the network, which can degrade the performance of the cache server network. The fault-tolerant caching algorithm proposed in this study ensures stable service quality by maintaining balance among media contents even when cache servers depart. Experimental results confirmed that the proposed algorithm effectively maintains content distribution despite the departure of cache servers. Additionally, we built a network composed of seven heterogeneous cache servers to verify the practicality of the proposed caching system and demonstrated its performance and scalability through a large-scale cache server network in a virtual environment.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.23 no.3
• /
• pp.119-125
• /
• 2023

IT infrastructure operation has advanced, and the methods for managing systems have become widely adopted. Recently, research has focused on improving system management using Syslog. However, utilizing log data collected through these methods presents challenges, as logs are extracted in various formats that require expert analysis. This paper proposes a system that utilizes edge computing to distribute the collection of Syslog data and preprocesses duplicate data before storing it in a central database. Additionally, the system constructs a data dictionary to classify and count data in real-time, with restrictions on transmitting registered data to the central database. This approach ensures the maintenance of predefined patterns in the data dictionary, controls duplicate data and temporal duplicates, and enables the storage of refined data in the central database, thereby securing fundamental data for big data analysis. The proposed algorithms and procedures are demonstrated through simulations and examples. Real syslog data, including extracted examples, is used to accurately extract necessary information from log data and verify the successful execution of the classification and storage processes. This system can serve as an efficient solution for collecting and managing log data in edge environments, offering potential benefits in terms of technology diffusion.

• Korean Journal of Psychosomatic Medicine
• /
• v.30 no.2
• /
• pp.172-178
• /
• 2022

Objectives : Although subclinical depression symptoms are associated with suicidal idea, most research have focused on clinical depression such as major depressive disorder or dysthymia. The aim of this study is to investigate network structure of depressive symptom and to reveal which symptoms are associated with suicidal ideation. Methods : We used part of data from the seventh Korea National Health and Nutrition Examination Survey. Participants were between 19 and 65 years of age (N=8,741). Network analysis with Isingfit model is used to reveal network structure of depressive symptoms and most central symptom and edges assessed by patient health questionnaire (PHQ-9). Results : The most two central symptoms were psychomotor activity and suicidal ideation. The strongest edge was psychomotor activity-suicidal ideation. Suicidal ideation also has strong association with depressive mood and worthlessness. Conclusions : These results suggest that psychomotor activity and suicidal ideation can serve as treatment target for subclinical depression and psychomotor activity, worthlessness and depressed mood may be important factor for early intervention of suicidal ideation.