cdeeparch a compact deep neural network architecture for
play

cDeepArch: A Compact Deep Neural Network Architecture for Mobile - PowerPoint PPT Presentation

Sep 16, 2022 •197 likes •437 views

cDeepArch: A Compact Deep Neural Network Architecture for Mobile Sensing Kang Yang 1 , Xiaoqing Gong 1 , Yang Liu 2 , Zhenjiang Li 2 , Tianzhang Xing 1 , Xiaojiang Chen 1 , Dingyi Fang 1 1 Northwest University, China 2 City University of Hong Kong

  1. cDeepArch: A Compact Deep Neural Network Architecture for Mobile Sensing Kang Yang 1 , Xiaoqing Gong 1 , Yang Liu 2 , Zhenjiang Li 2 , Tianzhang Xing 1 , Xiaojiang Chen 1 , Dingyi Fang 1 1 Northwest University, China 2 City University of Hong Kong 1

  2. Motivation … Camera + Learning Gyro. Technology Acc.

  3. Application ? Cognitive decline

  4. Application First-person view Cognitive aid system Recognizing pot cup open open close

  5. Common design principle . . . Rich sensor data Recognized by learning Applications

  6. Challenges Large targets . . .

  7. Challenges • Deep Learning Too large Big deep neural network Resource-limited

  8. Challenges • Deep Learning inaccurate Original Shrunk No quantitative measure on model model available resource conditions

  9. Any countermeasure? 0101… Server • Long and uncontrollable latency • High Service cost Potential privacy leakage •

  10. Our solution Context (office) Large targets . . .

  11. Our solution compact network (Office) Context recognition + Object recognition Context-oriented (computer, large and deep network target recognition mouse…) compact network adequate storage computation resource

  12. Our solution energy Context recognition computation + Available resource conditions Context-oriented target recognition • not based on designer’s experience • Formulation facilitated configuration

  13. Convolutional Neural Network Image data Conv1 Pool1 Conv2 Pool2 FC1 • Convolutional layer (dominant) • Pooling layer • Full connected layer

  14. Formulation facilitated configuration Selected W F W o *W o S P C (W+2P)*(W+2P)

  15. From computation to resource cost Conv1:64 fc:5 Conv2:128 computation resource(energy) designed network Conv1:16 Conv2:32 fc:5 : computation : actual resource consumption a small scale network

  16. Now… Context recognition + Object recognition Context-oriented target recognition Recognition task decomposition • • Formulation facilitated configuration From formulation to estimate the resource consumption •

  17. Enhancement: Convolutional layer Original Conv1 Conv2 Conv3 model Separated Conv1a Conv1b Conv2 Conv3 model #$ % − ⁄ 1 ) ⁄ ! ≤ ) 2

  20. Evaluation

  21. Experiments setup Dataset: • o Context recognition : § MIT Place2 (related to the daily contexts ) Object recognition : o § Cifar10 § Cifar100 (20 classes associated contexts)

  22. Evaluation results • Overall performance 10 targets 20 targets

  23. Conclusion 1, 2, 3 1. Large targets Decompose recognition task 2. Systematic way to configure network Execution overhead formulation facilitated configuration 3. Enhancement techniques

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend

Agenda Overview of deep learning Building a FAQ model with DeepLearning4J Integrating

Agenda Overview of deep learning Building a FAQ model with DeepLearning4J Integrating

Agenda Overview of deep learning Building a FAQ model with DeepLearning4J Integrating with a chatbot application Overview of deep learning AI ML Deep Learning Neural network architecture Neural network architecture Neural

469 views • 46 slides
A Neural Network Architecture for Detec2ng Gramma2cal Errors in SMT A Neural Network Architecture

A Neural Network Architecture for Detec2ng Gramma2cal Errors in SMT A Neural Network Architecture

A Neural Network Architecture for Detec2ng Gramma2cal Errors in SMT A Neural Network Architecture for Detec2ng Gramma2cal Errors in SMT Morpho-Syntac-c features outperform word embeddings on this task Syntac-c n-grams improve the

1.18k views • 38 slides
Deep Learning Primer Nishith Khandwala Neural Networks Overview Neural Network Basics

Deep Learning Primer Nishith Khandwala Neural Networks Overview Neural Network Basics

Deep Learning Primer Nishith Khandwala Neural Networks Overview Neural Network Basics Activation Functions Stochastic Gradient Descent (SGD) Regularization (Dropout) Training Tips and Tricks Neural Network (NN)

828 views • 45 slides
Neural Architecture Search Yu Cao What is Neural Architecture Search (NAS) Selecting the optimal

Neural Architecture Search Yu Cao What is Neural Architecture Search (NAS) Selecting the optimal

Neural Architecture Search Yu Cao What is Neural Architecture Search (NAS) Selecting the optimal network architecture automatically via machine instead of design it manually. It is an important aspect of AutoML. NAS search space 1.

3.43k views • 37 slides
Recurrent Neural Networks CS60010: Deep Learning Abir Das IIT Kharagpur Mar 11, 2020

Recurrent Neural Networks CS60010: Deep Learning Abir Das IIT Kharagpur Mar 11, 2020

Recurrent Neural Networks CS60010: Deep Learning Abir Das IIT Kharagpur Mar 11, 2020 Introduction Recurrent Neural Network LSTM Agenda Get introduced to different recurrent neural architecture e.g. , RNNs, LSTMs, GRUs etc. Get

743 views • 48 slides
Linearly Augmented Deep Neural Network Pegah Ghahremani, Johns Hopkins University, Jasha Droppo,

Linearly Augmented Deep Neural Network Pegah Ghahremani, Johns Hopkins University, Jasha Droppo,

Linearly Augmented Deep Neural Network Pegah Ghahremani, Johns Hopkins University, Jasha Droppo, Michael L. Seltzer, Microsoft Research Typical DNN Architecture Softmax L4 Layers are a composition of an affine and a non- linear function.

199 views • 19 slides
Deep Compression and EIE: Deep Neural Network Model Compression and Efficient Inference

Deep Compression and EIE: Deep Neural Network Model Compression and Efficient Inference

Deep Compression and EIE: Deep Neural Network Model Compression and Efficient Inference Engine Song Han CVA group, Stanford University Apr 7, 2016 1 Intro about me and my advisor Fourth year PhD with Prof. Bill Dally at Stanford.

2.06k views • 70 slides
Deep Learning (jkim@bi.snu.ac.kr) 2015/05/7

Deep Learning (jkim@bi.snu.ac.kr) 2015/05/7

Deep Learning (jkim@bi.snu.ac.kr) 2015/05/7 1 History of Neural Network Research Neural network Deep belief net Back propagation Science

1.27k views • 108 slides
Deep Compression and EIE: Deep Neural Network Model Compression and Efficient Inference

Deep Compression and EIE: Deep Neural Network Model Compression and Efficient Inference

Deep Compression and EIE: Deep Neural Network Model Compression and Efficient Inference Engine Song Han CVA group, Stanford University Jan 6, 2015 A few words about us Fourth year PhD with Prof. Bill Dally at Stanford.

966 views • 54 slides
Neural Information Retrieval Wassila Lalouani 1 Plan Neural network architectures Neural

Neural Information Retrieval Wassila Lalouani 1 Plan Neural network architectures Neural

Neural Information Retrieval Wassila Lalouani 1 Plan Neural network architectures Neural IR tasks Neural IR architecture Feature Representations Neural IR query auto completion Neural IR query suggestion Neural IR document

1.48k views • 18 slides
Bit Fu Bi Fusion on Bit-Level Dynamically Composable Architecture for Deep Neural Networks

Bit Fu Bi Fusion on Bit-Level Dynamically Composable Architecture for Deep Neural Networks

Bit Fu Bi Fusion on Bit-Level Dynamically Composable Architecture for Deep Neural Networks Hardik Sharma Georgia Institute of Technology Jongse Park Arm, Inc. Naveen Suda Liangzhen Lai University of California, San Diego

275 views • 24 slides
Concise Introduction to Deep Neural Networks Outline: Classification problems Motivating

Concise Introduction to Deep Neural Networks Outline: Classification problems Motivating

Concise Introduction to Deep Neural Networks Outline: Classification problems Motivating Deep (large) Neural Network (DNN) classifiers Neurons and DNN architectures Numerical training of DNNs (supervised deep learning) Spiking

630 views • 33 slides
Neural Networks: What can a network represent Deep Learning, Fall 2020 1 Recap : Neural

Neural Networks: What can a network represent Deep Learning, Fall 2020 1 Recap : Neural

Neural Networks: What can a network represent Deep Learning, Fall 2020 1 Recap : Neural networks have taken over AI Tasks that are made possible by NNs, aka deep learning Tasks that were once assumed to be purely in the human domain

1.92k views • 150 slides
ImageNet Classification with Deep Convolutional Neural Networks Krizhevsky et. all Outline

ImageNet Classification with Deep Convolutional Neural Networks Krizhevsky et. all Outline

ImageNet Classification with Deep Convolutional Neural Networks Krizhevsky et. all Outline Introduction DataSet Architecture of the Network Reducing overfitting Learning Results Discussion Introduction

1.3k views • 31 slides
ImageNet Classification with Deep Convolutional Neural Networks Alex Krizhevsky, Ilya Sutskever,

ImageNet Classification with Deep Convolutional Neural Networks Alex Krizhevsky, Ilya Sutskever,

ImageNet Classification with Deep Convolutional Neural Networks Alex Krizhevsky, Ilya Sutskever, Geoffrey E. Hinton Presented by Tugce Tasci, Kyunghee Kim 05/18/2015 Outline Goal DataSet Architecture of the Network Reducing

709 views • 35 slides
Propagating Error Backward Hyperparameters for Neural Networks } Multi-layer (deep) neural

Propagating Error Backward Hyperparameters for Neural Networks } Multi-layer (deep) neural

Learning in Neural Networks w 1,3 w 3,5 1 3 5 w w 1,4 3,6 w w 2,3 4,5 2 4 6 w w 2,4 4,6 Class #18: Back-Propagation; } A neural network can learn a classification function by Tuning Hyper-Parameters adjusting its weights to

259 views • 4 slides
Dominant Dimension and Orders over Cohen-Macaulay Rings 2. This Year 1. Last Year 1 Table of

Dominant Dimension and Orders over Cohen-Macaulay Rings 2. This Year 1. Last Year 1 Table of

Maurice Auslander Distinguished Lectures and International Conference zgr Esentepe April 25, 2019 University of Toronto Dominant Dimension and Orders over Cohen-Macaulay Rings 2. This Year 1. Last Year 1 Table of contents Last Year

733 views • 17 slides
Speech recognition in the presence of highly non-stationary noise based on spatial, spectral and

Speech recognition in the presence of highly non-stationary noise based on spatial, spectral and

Speech recognition in the presence of highly non-stationary noise based on spatial, spectral and temporal speech/ noise modeling combined with dynamic variance adaptation M. Delcroix, K. Kinoshita, T. Nakatani, S. Araki, A. Ogawa, T.

422 views • 26 slides
The The W World Oce ld Ocean an Re Regime: Polluted Wa Waters and Exhausted Fi Fishe

The The W World Oce ld Ocean an Re Regime: Polluted Wa Waters and Exhausted Fi Fishe

The The W World Oce ld Ocean an Re Regime: Polluted Wa Waters and Exhausted Fi Fishe heries Peter J. Jacques, Professor School of Politics, Security, and International Affairs & The National Center for Integrated Coastal Research

279 views • 16 slides
GPSCP: A General-Purpose Support-Circuit Preconditioning Approach to Large Scale SPICE Accurate

GPSCP: A General-Purpose Support-Circuit Preconditioning Approach to Large Scale SPICE Accurate

Design Automation Group GPSCP: A General-Purpose Support-Circuit Preconditioning Approach to Large Scale SPICE Accurate Preconditioning Approach to Large-Scale SPICE-Accurate Nonlinear Circuit Simulations Authors: Xueqian Zhao Xueqian Zhao

390 views • 25 slides
Germany: The Dominant Power in Europe The Dominant Power in Europe meelis_kitsing@uml.edu

Germany: The Dominant Power in Europe The Dominant Power in Europe meelis_kitsing@uml.edu

Germany: The Dominant Power in Europe The Dominant Power in Europe meelis_kitsing@uml.edu Source: Palmer 2004 Source: Palmer 2004 Population: 82,431,390 (July 2005 est.) Life expectancy at birth: 78.65 years (total population) 75.66 years

689 views • 55 slides
Tractable Term Structure ModelsA New Approach Bruno Feunou, Jean-S ebastien Fontaine, Anh

Tractable Term Structure ModelsA New Approach Bruno Feunou, Jean-S ebastien Fontaine, Anh

Tractable Term Structure ModelsA New Approach Bruno Feunou, Jean-S ebastien Fontaine, Anh Le, Christian Lundblad Bank of Canada and FRBSF Fixed Income Conference November 2015 Feunou, Fontaine, Le Tractable Term Structure Modeling: A New

648 views • 37 slides
Investment Funds and the Corporate-Bond Market in the Covid-19 Crisis Itay Goldstein Wharton

Investment Funds and the Corporate-Bond Market in the Covid-19 Crisis Itay Goldstein Wharton

Investment Funds and the Corporate-Bond Market in the Covid-19 Crisis Itay Goldstein Wharton School, University of Pennsylvania Hutchins Center, Brookings Institution May 27, 2020 Itay Goldstein: Corporate-Bond Market in the Covid-19 Crisis

290 views • 12 slides
Vi Viden en: A Attack cker er Id Iden entifi fication on on on In In-V -Veh ehicl cle

Vi Viden en: A Attack cker er Id Iden entifi fication on on on In In-V -Veh ehicl cle

Vi Viden en: A Attack cker er Id Iden entifi fication on on on In In-V -Veh ehicl cle Ne e Networ orks Kyong-Tak Cho and Kang G. Shin Univ Univer ersity o sity of Mic f Michig higan, A an, Ann A nn Arbo rbor r CCS 2017

253 views • 24 slides

More recommend