Sparsification and Separation of Deep Learning Layers for Constrained Resource Inference ...

by Jinse Kwon on 2018-03-30 16:04:38

Date : 2018. 04. 04 (Wed) 13:00 Locate : EB5. 533 Presenter : Jinse Kwon   Title : Sparsification and Separation of Deep Learning Layers for Constrained Resource Inference on Wearables Author : Sourav Bhattacharya, Nicholas D. Lane (Nokia Bell Labs and University College London)   Abstract : Deep learning has revolutionized the way sensor data are analyzed and interpreted. The accuracy gains these approaches offer make them attractive for the next generation of mobile, wearable and embedded sensory applications. However, state-of-the-art deep learning algorithms typically require a significant amount of device and processor resources, even just for the inference stages that are used to discriminate high-level classes from low-level data. The limited availability of memory, computation, and energy on mobile and embedded platforms thus pose a significant challenge to the adoption of these powerful learning techniques. In this paper, we ... Continue reading →

77 Views

리눅스 기반 모바일 기기에서 사용자 응답성 향상을 위한 프레임워크 지원 선별적 페이지 보호 기법

by Byungkyo Jung on 2018-03-14 15:48:47

Data : 2018.3.21 (Wed) 13:00 Locate : EB5. 533 Presenter : Byeongkyo Cheong Author : 김승준, 김정호, 홍성수 Abstract : While Linux-based mobile devices such as smartphones are increasingly used, they often exhibit poor response time. One of the factors that influence the user-perceived interactivity is the high page fault rate of interactive tasks. Pages owned by interactive tasks can be removed from the main memory due to the memory contention between interactive and background tasks. Since this increases the page fault rate of the interactive tasks, their executions tend to suffer from increased delays. This paper proposes a framework-assisted selective page protection mechanism for improving interactivity of Linux-based mobile devices. The framework-assisted selective page protection enables the run-time system to identify interactive tasks at the framework level and to deliver their IDs to the kernel. As a result, the kernel can maintain the pages owned by the identified ... Continue reading →

61 Views

Predicting Latent Narrative Mood Using Audio and Physiologic Data

by Jinyoung Choi on 2018-03-11 11:05:37

Date : 2018. 3. 13 (Wed) 13:00 Locate : EB5. 533 Presenter : Jinyoung Choi   Title : Predicting Latent Narrative Mood Using Audio and Physiologic Data AAAI 2017, 948-954 Author : Tuka Waddah AlHanai, Mohammad MAhdi Ghassemi, Massachusetts Institute of Techonology, Cambridge MA, USA Abstract : Inferring the latent emotive content of a narrative requires consideration of para-linguistic cues (e.g. pitch), linguistic content (e.g. vocabulary) and the physiological state of the narrator (e.g. heart-rate). In this study we utilized a combination of auditory, text, and physiological signals to predict the mood (happy or sad) of 31 narrations from subjects engaged in personal story-telling. We extracted 386 audio and 222 physiological features (using the Samsung Simband) from the data. A subset of 4 audio , 1 text, and 5 physiologic features were identified using Sequential Forward Selection (SFS) for inclusion in a Neural Network (NN). These features ... Continue reading →

112 Views

MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications

by Do Trung Hai on 2018-02-05 11:28:34

Date : 2018. 02. 05 (Mon) 17:00 Locate : EB5. 533 Presenter : Trunghai Do Author : Andrew G. Howard, Menglong Zhu, Bo Chen, Dmitry Kalenichenko, Weijun Wang, Tobias Weyand, Marco Andreetto, Hartwig Adam  Abstract : We present a class of efficient models called MobileNets for mobile and embedded vision applications. MobileNets are based on a streamlined architecture that uses depthwise separable convolutions to build light weight deep neural networks. We introduce two simple global hyperparameters that efficiently trade off between latency and accuracy. These hyper-parameters allow the model builder to choose the right sized model for their application based on the constraints of the problem. We present extensive experiments on resource and accuracy tradeoffs and show strong performance compared to other popular models on ImageNet classification. We then demonstrate the effectiveness of MobileNets across a wide ... Continue reading →

121 Views

The Performance of μ-Kernel-Based Systems

by Daeyoung Song on 2018-01-12 17:45:41

Title : The Performance of μ-Kernel-Based Systems SOSP(Symposium on Operating Systems Principles) ‘97 Author : Hermann H¨artig, Michael Hohmuth, Sebastian Sch¨onberg Jean Wolter (Department of Computer Science, Dresden University of Technology, Germany), Jochen Liedtke (IBM T.J. Watson Research Center)  Abstract : First-generation p-kernels have a reputation for being too slow and lacking sufficient flexibility. To determine whether LA, a lean second generation p-kernel, has overcome these limitations, we have repeated several earlier experiments and conducted some novel ones. Moreover, we  ported the Linux operating system to run on top of the L4 p-kernel and compared the resulting system with both Linux running native, and MkLinux, a Linux version that executes on top of a first-generation Mach derived p-kernel. For L4Linux, the AIM benchmarks report a maximum\ throughput which is only 5% lower than that of native Linux. The  ... Continue reading →

124 Views

HiCH: Hierarchical Fog-Assisted Computing Architecture for Healthcare IoT

by Sihyeong Park on 2017-11-14 17:37:30

Date : 2017. 11. 20 (Mon) 17:00 Locate : EB5. 533 Presenter : Sihyeong Park   Title : HiCH: Hierarchical Fog-Assisted Computing Architecture for Healthcare IoT ACM Transactions on Embedded Computing Systems (TECS) - Special Issue ESWEEK 2017, CASES 2017, CODES + ISSS 2017 and EMSOFT 2017 TECS  Volume 16 Issue 5s, October 2017  Article No. 174    Author : Iman Azimi University of Turku, Turku, Finland, Arman Anzanpour University of Turku, Turku, Finland, Amir M. Rahmani University of California Irvine and TU Wien, Irvine, CA, USA, Tapio Pahikkala University of Turku, Turku, Finland, Marco Levorato University of California Irvine, Irvine, CA, USA, Pasi Liljeberg University of Turku, Turku, Finland, Nikil Dutt University of California Irvine, Irvine, CA, US Abstract : The Internet of Things (IoT) paradigm holds significant promises for remote health monitoring systems. Due to their life- or mission-critical nature, these ... Continue reading →

257 Views

BackDoor: Making Microphones Hear Inaudible Sounds

by Jinse Kwon on 2017-11-09 15:32:30

Date : 2017. 11. 13 (Mon) 17:00 Locate : EB5. 533 Presenter : Jinse Kwon   Title : BackDoor: Making Microphones Hear Inaudible Sounds (mobisys 2017 Best paper) Author : Nirupam Roy, Haitham Hassanieh, Romit Roy Choudhury (University of Illinois at Urbana-Champaign)   Abstract : Consider sounds, say at 40kHz, that are completely outside the human's audible range (20kHz), as well as a microphone's recordable range (24kHz). We show that these high frequency sounds can be designed to become recordable by unmodified microphones, while remaining inaudible to humans. The core idea lies in exploiting non-linearities in microphone hardware. Briefly, we design the sound and play it on a speaker such that, after passing through the microphone's non-linear diaphragm and power-amplifier, the signal creates a "shadow" in the audible frequency range. The shadow can be regulated to carry data bits, thereby enabling an acoustic ... Continue reading →

259 Views

YOLO-based Object Detection on ARM Mali GPU

by Do Trung Hai on 2017-11-04 15:06:05

Date : 2017.11.06(Mon) 05:00 P.M. Locate : EB5. 533 Presenter : Trunghai Do Title : YOLO-based Object Detection on ARM Mali GPU Author : Trunghai Do, Jemin Lee, Hyungshin Kim Abstract Nowadays, convolutional neural networks become the heart of many computer vision solutions to solve a wide range of tasks including image classification, object detection and  segmentation. In this paper, we present YOLO implementation on ARM Mali-T628 MP6 GPU of ODROID-XU4. Original YOLO algorithm was implemented by Darknet framework dedicated for NVIDIA GPU. Unlike Darknet, we use CK-Caffe as a deep learning backend to run YOLO on ARM-based GPUs. Additionally, we replace all fully connected layers which act as detection layers in YOLO with convolutional layers to reduce the model size. We train our GoogLeNet-based and Fast YOLO-based models on canonical PASCAL VOC2007 + VOC2012 "trainval" (training + validation) set on NVIDIA GTX-1080 and experiment on ... Continue reading →

245 Views

스마트폰 알림 장소에 따른 사용자 반응 분석

by Seula Hwang on 2017-10-30 14:06:01

Date : 2017.10.30(Mon) 05:30 P.M. Locate : EB5. 533 Presenter : Hwang, Seula Title : 스마트폰 알림 장소에 따른 사용자 반응 분석 Article source:  Continue reading →

173 Views

더블 버퍼링을 사용하는 OpenCL 프로그램에서 글로벌 워크 크기 최적화

by Ikhee Shin on 2017-10-30 13:21:21

  Date : 2017.10.30(Mon) 05:00 P.M. Locate : EB5. 533 Presenter : Ikhee Shin Title : 더블 버퍼링을 사용하는 OpenCL 프로그램에서 글로벌 워크 크기 최적화 Continue reading →

150 Views

임베디드 하드웨어 저전력 기능을 활용한 웨어러블 운영체제의 하이브리드 가버너

by Sungyup Lee on 2017-10-22 20:32:05

Date : 2017.10.22(Mon) 05:00 P.M. Locate : EB5. 533 Presenter : Sungyup Lee Title : 임베디드 하드웨어 저전력 기능을 활용한 웨어러블 운영체제의 하이브리드 가버너 Continue reading →

110 Views

No Need to Stop What You’re Doing: Exploring No-Handed Smartwatch Interaction

by Jinyoung Choi on 2017-06-23 19:23:54

Date : 2017.06.26(Mon) 04:00 P.M. Locate : EB5. 533 Presenter : Jinyoung Choi Title : No Need to Stop What You’re Doing: Exploring No-Handed Smartwatch Interaction Author : Seongkook Heo, Michelle Annett, Ben Lafreniere, Tovi Grossman, George Fitzmaurice Abstract Smartwatches have the potential to enable quick micro-interactions throughout daily life. However, because they require both hands to operate, their full potential is constrained, particularly in situations where the user is actively performing a task with their hands. We investigate the space of no-handed interaction with smartwatches in scenarios where one or both hands are not free. Specifically, we present a taxonomy of scenarios in which standard touchscreen interaction with smartwatches is not possible, and discuss the key constraints that limit such interaction. We then implement a set of interaction techniques and evaluate them via two user studies: one where participants viewed video ... Continue reading →

252 Views

Performance Analysis of IEEE 802.15.4 Non-beacon Mode with the Unslotted CSMA/CA

by Sungyup Lee on 2017-05-15 11:23:24

Date : 2017.05.15(Mon) 04:00 P.M. Locate : EB5. 533 Presenter : Sungyup Lee Title : Performance Analysis of IEEE 802.15.4 Non-beacon Mode with the Unslotted CSMA/CA Author : Tae Ok Kim, Jin Soo Park, Hak Jin Chong, Kyung Jae Kim, Bong Dae Choi, Member of IEEE Abstract We provide a simple mathematical model of IEEE 802.15.4 unslotted CSMA/CA by busy cycle of M/G/1 queueing system to obtain several performance measures. Our performance measures can be used for determining the optimal number of devices while supporting the required QoS constraints on the average packet delay and the packet loss probability, and also for calculating battery lifetime. Proceeding : IEEE COMMUNICATIONS LETTERS, VOL. 12, NO.4, APRIL 2008 Continue reading →

327 Views

Adaptive Optimization for OpenCL Programs on Embedded Heterogeneous Systems

by Ikhee Shin on 2017-05-08 09:07:26

Date : 2017.05.08(Mon) 04:00 P.M. Locate : EB5. 533 Presenter : Ikhee Shin Title : Adaptive Optimization for OpenCL Programs on Embedded Heterogeneous Systems Author : Ben Taylor, Vicent Sanz Marco, Zheng Wang  Abstract Heterogeneous multi-core architectures consisting of CPUs and GPUs are commonplace in today’s embedded systems. These architectures offer potential for energy efficient computing if the application task is mapped to the right core. Realizing such potential is challenging due to the complex and evolving natural of hardware and applications. This paper presents an automatic approach to map OpenCL kernels onto heterogeneous multi-cores for a given optimization criterion – whether it is faster runtime, lower energy consumption or a trade-off between them. This is achieved by developing a machine learning based approach to predict which processor to use to run the OpenCL kernel and the host program, and at what frequency ... Continue reading →

412 Views

Predicting Sensory Data and Extending Battery Life for Wearable Devices

by Seula Hwang on 2017-04-28 14:38:15

Date : 2017.05.01(Mon) 04:00 P.M. Locate : EB5. 533 Presenter : Seula Hwang Title : Predicting Sensory Data and Extending Battery Life for Wearable Devices Author : Songchun Fan, Qiuyun Llull, Benjamin C. Lee Abstract Telepath is a framework that supports communication-free offloading for wearable devices. With offline training, activity recognition tasks can be offloaded from the wearable to the user's phone, without transferring raw sensing data. The key observation is that when the user is carrying both devices, the sensing streams on the two devices are highly correlated. By exploiting the correlation, the phone can estimate the wearable's sensing data and emulate the watch. Our evaluations shows that with Telepath, the phone performs accurately on activity recognition tasks that are designed for smart watches, achieving on average 87% of the watch's accuracy while extending the watch's battery life by 2.1x. Proceeding ... Continue reading →

421 Views