A 28nm SoC with a 1.2GHz 568nJ/ Prediction Sparse Deep-Neural-Network Engine with >0.1 Timing Error Rate Tolerance for IoT Applications
|
|
- Jason Carroll
- 5 years ago
- Views:
Transcription
1 A 28nm SoC with a 1.2GHz 568nJ/ Prediction Sparse Deep-Neural-Network Engine with >0.1 Timing Error Rate Tolerance for IoT Applications Paul Whatmough, S. K. Lee, H. Lee, S. Rama, D. Brooks, G.-Y. Wei Harvard University, Cambridge, MA 1 of 86
2 Motivation 2 of 86
3 Motivation Deep Neural Networks (DNNs) A universal classifier with many diverse applica5ons Interpret noisy real- world data from sensor- rich systems 3 of 86
4 Motivation Deep Neural Networks (DNNs) A universal classifier with many diverse applica5ons Interpret noisy real- world data from sensor- rich systems Large memory footprint and compute load Perform DNN inference on device in micro- Joules 4 of 86
5 Inputs Motivation DNN ENGINE Outputs Sensor Data + DNN Topology + Weights Input Vector Weight Neuron Hidden Layers Output Classes Argmax Classification DNN ENGINE A Shared Programmable Classifier Specialized Architecture and Efficient Digital Circuits 5 of 86
6 Inputs Motivation DNN ENGINE Outputs Sensor Data + DNN Topology + Weights Input Vector Weight Neuron Hidden Layers Output Classes Argmax Classification DNN ENGINE A Shared Programmable Classifier Specialized Architecture and Efficient Digital Circuits Parallelism: Balanced memory bandwidth and throughput 6 of 86
7 Inputs Motivation DNN ENGINE Outputs Sensor Data + DNN Topology + Weights Input Vector Weight Neuron Hidden Layers Output Classes Argmax Classification DNN ENGINE A Shared Programmable Classifier Specialized Architecture and Efficient Digital Circuits Parallelism: Balanced memory bandwidth and throughput Sparsity: HW support for sparse data and small datatypes 7 of 86
8 Inputs Motivation DNN ENGINE Outputs Sensor Data + DNN Topology + Weights Input Vector Weight Neuron Hidden Layers Output Classes Argmax Classification DNN ENGINE A Shared Programmable Classifier Specialized Architecture and Efficient Digital Circuits Parallelism: Balanced memory bandwidth and throughput Sparsity: HW support for sparse data and small datatypes Resilience: Razor adapta5on to minimize PVT margins 8 of 86
9 Outline Mo5va5on DNN Engine Parallelism Sparsity Resilience Measurement Results Summary 9 of 86
10 Fully-Connected DNN Graph Input Vector Output Classes Input Layer Output Layer Hidden Layers 10 of 86
11 Fully-Connected DNN Graph Weight Neuron Input Vector Output Classes 11 of 86
12 Fully-Connected DNN Graph SIMD Window Input Vector Output Classes 12 of 86
13 Fully-Connected DNN Graph Weights Input Vector Output Classes 13 of 86
14 Fully-Connected DNN Graph Input Vector Output Classes 14 of 86
15 Fully-Connected DNN Graph Input Vector Output Classes 15 of 86
16 Fully-Connected DNN Graph Input Vector Output Classes 16 of 86
17 Fully-Connected DNN Graph Input Vector Output Classes 17 of 86
18 Fully-Connected DNN Graph Input Vector Output Classes 18 of 86
19 Fully-Connected DNN Graph Input Vector Output Classes 19 of 86
20 Fully-Connected DNN Graph Input Vector Output Classes 20 of 86
21 Fully-Connected DNN Graph Input Vector Output Classes 21 of 86
22 Balancing Efficiency and Bandwidth Data"Reads"(Norm.)" 1" 0.8" 0.6" 0.4" 0.2" 0" No Data Reuse 1" 4" 8" 16" 32" SIMD"Width"(Words)" Parallelism increases throughput and data reuse 30" 20" 10" 0" Weight"Reads"/"Cycle" 22 of 86
23 Balancing Efficiency and Bandwidth Data"Reads"(Norm.)" 1" 0.8" 0.6" 0.4" 0.2" 0" No Data Reuse Bandwidth too high 1" 4" 8" 16" 32" SIMD"Width"(Words)" Parallelism increases throughput and data reuse But also increases Memory Bandwidth demands 30" 20" 10" 0" Weight"Reads"/"Cycle" 23 of 86
24 Balancing Efficiency and Bandwidth Data"Reads"(Norm.)" 1" 0.8" 0.6" 0.4" 0.2" 0" No Data Reuse 1" 4" 8" 16" 32" SIMD"Width"(Words)" Parallelism increases throughput and data reuse But also increases Memory Bandwidth demands 128b/cycle Bandwidth too high 8- Way SIMD is efficient at reasonable memory BW 10x Ac5va5on reuse, with 128b AXI channel 30" 20" 10" 0" Weight"Reads"/"Cycle" 24 of 86
25 DNN ENGINE Micro-Architecture Host Processor Data In CFG Data Out Data Read IPBUF XBUF Writeback NBUF W- MEM DMA Weight Load MAC Datapath Activation 25 of 86
26 DNN ENGINE Micro-Architecture Host Processor Data In CFG Data Out Data Read IPBUF XBUF Writeback NBUF W- MEM DMA Weight Load MAC Datapath Activation Host Processor loads configura5on and input data 26 of 86
27 DNN ENGINE Micro-Architecture Host Processor Data In CFG Data Out Data Read IPBUF XBUF Writeback NBUF W- MEM DMA Weight Load MAC Datapath Activation Accumulate products of Ac5va5on and Weights 27 of 86
28 DNN ENGINE Micro-Architecture Host Processor Data In CFG Data Out Data Read IPBUF XBUF Writeback NBUF W- MEM DMA Weight Load MAC Datapath Activation Add bias, apply ReLU ac5va5on and writeback 28 of 86
29 DNN ENGINE Micro-Architecture Host Processor Data In CFG Data Out Data Read IPBUF XBUF Writeback NBUF W- MEM DMA Weight Load MAC Datapath Activation IRQ to host, which retrieves predic5on data 29 of 86
30 Outline Mo5va5on DNN Engine Parallelism Sparsity Resilience Measurement Results Summary 30 of 86
31 Exploiting Sparse Data in DNNs Input Vector Output Classes 31 of 86
32 Exploiting Sparse Data in DNNs Input Vector Output Classes 32 of 86
33 Exploiting Sparse Data in DNNs Input Vector Output Classes 33 of 86
34 Exploiting Sparse Data in DNNs Input Vector Output Classes 34 of 86
35 Exploiting Sparse Data in DNNs Input Vector Output Classes 35 of 86
36 Exploiting Sparse Data in DNNs Input Vector Output Classes 36 of 86
37 Exploiting Sparse Data in DNNs Input Vector Output Classes 37 of 86
38 Exploiting Sparse Data in DNNs Input Vector Output Classes 38 of 86
39 Exploiting Sparse Data in DNNs Input Vector Output Classes 39 of 86
40 Exploiting Sparse Data in DNNs Input Vector Output Classes 40 of 86
41 Exploiting Sparse Data in DNNs Input Vector Output Classes 41 of 86
42 Exploiting Sparse Data in DNNs Input Vector Output Classes 42 of 86
43 DNN ENGINE Micro-Architecture Host Processor Data In CFG Data Out Data Read IPBUF XBUF Writeback NBUF W- MEM DMA Weight Load MAC Datapath Activation 43 of 86
44 DNN ENGINE Micro-Architecture Host Processor Data In CFG Data Out Data Read IPBUF XBUF Writeback NBUF W- MEM SKIP Index DMA Weight Load MAC Datapath Activation 44 of 86
45 Outline Mo5va5on DNN Engine Parallelism Sparsity Resilience Measurement Results Summary 45 of 86
46 Razor Hardware Accelerators CLK process safety voltage temp coupling jizer ageing 46 of 86
47 Razor Hardware Accelerators CLK process safety voltage temp coupling jizer ageing Occasionally allow real cri5cal paths to fail 5ming Explicitly check for late- arriving signals Adapt Vdd/Fclk to track near- zero error- rate 47 of 86
48 Razor Hardware Accelerators CLK process safety voltage temp coupling jizer ageing Occasionally allow real cri5cal paths to fail 5ming Explicitly check for late- arriving signals Adapt Vdd/Fclk to track near- zero error- rate Must ensure 5ming errors do not affect system Roll- back and replay [Das et al., CICC 13] Algorithmic error tolerance [Whatmough et al., ISSCC 13] 48 of 86
49 DNN Engine Micro-Architecture Host Processor Data In CFG Data Out Data Read IPBUF XBUF RZ Writeback NBUF W- MEM RZ SKIP Index DMA Weight Load MAC Datapath Activation Two Razor stages: W Memory and MAC Datapath 49 of 86
50 Memory: Algorithm-Level Tolerance Count%(a.u.)% Two s Compliment (TC) -1: : !0.3%!0.2%!0.1% 0.0% 0.1% 0.2% 0.3% Weight%Value% Weights are zero- mean Gaussians (regulariza5on) Small magnitude, random sign (worst- case in TC) 50 of 86
51 Memory: Algorithm-Level Tolerance Count%(a.u.)%!0.3%!0.2%!0.1% 0.0% 0.1% 0.2% 0.3% Weight%Value% Two s Compliment (TC) -1: : Sign- Magnitude (SM) -1: : Weights are zero- mean Gaussians (regulariza5on) Small magnitude, random sign (worst- case in TC) Sign- Magnitude (SM) reduces switching in weights Lowers switched cap (power) and bit error rate 51 of 86
52 Datapath: Circuit-Level Tolerance X RZ Y W RZ Algorithmic error tolerance in datapath is poor Errors are persistent in accumulator Circuit- level error tolerance using 5me borrowing Single- sided cri5cal path at accumulator register 52 of 86
53 Datapath: Circuit-Level Tolerance X W RZ Cri5cal Path RZ Y Algorithmic error tolerance in datapath is poor Errors are persistent in accumulator Circuit- level error tolerance using 5me borrowing Single- sided cri5cal path at accumulator register 53 of 86
54 Datapath: Circuit-Level Tolerance Slack X W RZ Cri5cal Path RZ Y Algorithmic error tolerance in datapath is poor Errors are persistent in accumulator Circuit- level error tolerance using 5me borrowing Single- sided cri5cal path at accumulator register 54 of 86
55 Outline Mo5va5on DNN Engine Parallelism Sparsity Resilience Measurement Results Summary 55 of 86
56 28nm SoC for IoT Applications 28nm SoC Test Chip Cortex- M0 Subsystem Accelerator Subsystem OSC HCLK ARM Cortex- M0 M S V DCO DCO FCLK USB Scan GPIO RTC USB Low- BW Peripherals Timers Wdog UARTs BIST S S S V SOC 32b APB UART SCAN GPIO SYSCTL I- MEM 64KB D- MEM 64KB Bridge S M S M M S S S S V SOC 32b AHB S ASYNC DNN Engine W- MEM 1MB M V ACC!Accelerator Logic V MEMP!SRAM Periphery V MEMC!SRAM Core S 128b AXI 56 of 86
57 Razor Voltage Scaling V DD Scaling (F CLK = 667MHz) Timing Violation Rate TC SM Sign- off Power (mw) FC- DNN (98.36% MNIST) 57 of 86
58 Razor Voltage Scaling V DD Scaling (F CLK = 667MHz) Timing Violation Rate PoFF 30% Power Reduc5on TC SM Sign- off Power (mw) Opera5on at PoFF (770mV) lowers power by 30% 58 of 86
59 Razor Voltage Scaling V DD Scaling (F CLK = 667MHz) Timing Violation Rate Zero Margin Margin PoFF 30% Power Reduc5on TC SM Sign- off Power (mw) Margin for fast varia5ons down to 715mV 59 of 86
60 Timing Error Tolerance Summary Timing Violation Rate % Accuracy Memory Datapath Combined > 1,000,000 X TC SM BM TC SM TB TC ALL Aggregate 5ming viola5on rate greater than of 86
61 Normalized Energy/Pred Energy/Throughput Summary 98.36% Accuracy - 30% Energy Throughput Normalized Pred/Sec TC SM 0.9V 667MHz Sign-off 8b 61 of 86
62 Normalized Energy/Pred Energy/Throughput Summary 98.36% Accuracy - 30% - 30% +4X - 4X Energy Throughput Normalized Pred/Sec TC SM 8b AVS SKIP 0.9V 667MHz Sign-off 0.77V 667MHz PoFF 62 of 86
63 Normalized Energy/Pred Energy/Throughput Summary 98.36% Accuracy - 30% - 30% +4X - 4X Energy Throughput 260nJ Normalized Pred/Sec TC SM 8b AVS SKIP AVS 0.9V 667MHz Sign-off 0.77V 667MHz PoFF 0.715V 667MHz 63 of 86
64 Energy/Throughput Summary Normalized Energy/Pred % Accuracy Energy Throughput - 30% +50% 260nJ +4X - 4X Normalized Pred/Sec TC SM 8b AVS SKIP AVS AFS SKIP 0.9V 667MHz Sign-off 0.77V 667MHz PoFF 0.715V 667MHz 0.9V 1GHz PoFF 64 of 86
65 Energy/Throughput Summary Normalized Energy/Pred % Accuracy Energy Throughput - 30% +50% 260nJ +4X - 4X +20% 570nJ Normalized Pred/Sec TC SM 8b AVS SKIP AVS AFS SKIP AFS 0.9V 667MHz Sign-off 0.77V 667MHz PoFF 0.715V 667MHz 0.9V 1GHz PoFF 0.9V 1.2GHz 65 of 86
66 Summary and Conclusion DNN Engine A Shared Programmable Classifier Parallelism: 10X Data 128b/cycle BW Sparsity: +4X Throughput, - 4X Energy Resilience: +50% Throughput or - 30% Energy w/razor Energy Efficient DNN GHz Timing Error Tolerance Pe 98.36% (MNIST) International Solid-State Circuits Conference with >0.1 Timing Error Rate Tolerance for IoT Applications 66 of 86
67 67 of 86
Neural Networks II. Chen Gao. Virginia Tech Spring 2019 ECE-5424G / CS-5824
Neural Networks II Chen Gao ECE-5424G / CS-5824 Virginia Tech Spring 2019 Neural Networks Origins: Algorithms that try to mimic the brain. What is this? A single neuron in the brain Input Output Slide
More informationCloud, Distributed, Embedded. Erlang in the Heterogeneous Computing World. Omer
Cloud, Distributed, Embedded. Erlang in the Heterogeneous Computing World Omer Kilic @OmerK omer@erlang-solutions.com Outline Challenges in modern computing systems Heterogeneous computing Co-processors
More informationFast Floating Point Compression on the Cell BE Processor
Fast Floating Point Compression on the Cell BE Processor Ajith Padyana, T.V. Siva Kumar, P.K.Baruah Sri Satya Sai University Prasanthi Nilayam - 515134 Andhra Pradhesh, India ajith.padyana@gmail.com, tvsivakumar@gmail.com,
More informationImplementation of Height Measurement System Based on Pressure Sensor BMP085
017 nd International Conference on Test, Measurement and Computational Method (TMCM 017) ISBN: 978-1-60595-465-3 Implementation of Height Measurement System Based on Pressure Sensor BMP085 Gao-ping LIU
More informationComputing s Energy Problem:
Computing s Energy Problem: (and what we can do about it) Mark Horowitz Stanford University horowitz@ee.stanford.edu 1 of 46 Everything Has A Computer Inside 2of 46 The Reason is Simple: Moore s Law Made
More informationSafety Critical Systems
Safety Critical Systems Mostly from: Douglass, Doing Hard Time, developing Real-Time Systems with UML, Objects, Frameworks And Patterns, Addison-Wesley. ISBN 0-201-49837-5 1 Definitions channel a set of
More informationAttacking and defending neural networks. HU Xiaolin ( 胡晓林 ) Department of Computer Science and Technology Tsinghua University, Beijing, China
Attacking and defending neural networks HU Xiaolin ( 胡晓林 ) Department of Computer Science and Technology Tsinghua University, Beijing, China Outline Background Attacking methods Defending methods 2 AI
More informationACCURATE PRESSURE MEASUREMENT FOR STEAM TURBINE PERFORMANCE TESTING
ACCURATE PRESSURE MEASUREMENT FOR STEAM TURBINE PERFORMANCE TESTING Blair Chalpin Charles A. Matthews Mechanical Design Engineer Product Support Manager Scanivalve Corp Scanivalve Corp Liberty Lake, WA
More informationVLSI Design 12. Design Styles
1 9 8 5 1 9 8 7 1 9 9 1 1 9 9 3 1 9 9 9 2 0 0 1 2 0 0 5 2 0 0 7 2 0 0 9 Pr odu ctiv it y (T r an s./ St a f -M o nt h ) VLSI Design Last module: Floorplanning Sequential circuit design Clock skew This
More informationNEURAL NETWORKS BASED TYRE IDENTIFICATION FOR A TYRE INFLATOR OPERATIONS
Lfe/sou/n /oh NEURAL NETWORKS BASED TYRE IDENTIFICATION FOR A TYRE INFLATOR OPERATIONS A Thesis submitted to the Department of Electrical Engineering, University of Moratuwa On partial fulfilment of the
More informationFunctional safety. Functional safety of Programmable systems, devices & components: Requirements from global & national standards
Functional safety Functional safety of Programmable systems, devices & components: Requirements from global & national standards Matthias R. Heinze Vice President Engineering TUV Rheinland of N.A. Email
More informationFlight Software Overview
Solar Probe Plus A NASA Mission to Touch the Sun Flight Software Overview Christopher Krupiarz Christopher.Krupiarz@jhuapl.edu Overarching Science Objective To determine the structure and dynamics of the
More informationDesign of AMBA APB Protocol
Design of AMBA APB Protocol G. Venkat Rao 1, Pasam Abhigna 2, J. Naveen 3, S. Vineeth Reddy 4 1 Assistant Professor Department of ECE Geethanjali College of engineering and technology 2, 3, 4 Bachelor
More informationIntroduction to Machine Learning NPFL 054
Introduction to Machine Learning NPFL 054 http://ufal.mff.cuni.cz/course/npfl054 Barbora Hladká hladka@ufal.mff.cuni.cz Martin Holub holub@ufal.mff.cuni.cz Charles University, Faculty of Mathematics and
More informationNeural Network in Computer Vision for RoboCup Middle Size League
Journal of Software Engineering and Applications, 2016, *,** Neural Network in Computer Vision for RoboCup Middle Size League Paulo Rogério de Almeida Ribeiro 1, Gil Lopes 1, Fernando Ribeiro 1 1 Department
More informationSurfing Interconnect
Surfing Interconnect Mark R. Greenstreet and Jihong Ren University of British Columbia, Rambus Surfing Interconnect p.1/17 Motivation Wires are the problem: Wires scale poorly with feature size: Gate delays
More informationsecuring networks with silicon
NITROX II A Family of In-line Security Processors Presented by: Muhammad Raghib Hussain Contributed by: Bob Sanzone, Dan Katz, David Asher, David Carlson, Gregg Bouchard, Michael Bertone, Muhammad Hussain,
More informationOxygen Meter User Manual
Oxygen Meter User Manual Monday, July 23, 2007 1. Outline...2 2. Program...3 2.1. Environment for program execution...3 2.2. Installation...3 2.3. Un installation...3 2.4. USB driver installation...3 2.5.
More informationThe Cooperative Cleaners Case Study: Modelling and Analysis in Real-Time ABS
: Modelling and Analysis in Real-Time ABS Silvia Lizeth Tapia Tarifa Precise Modelling and Analysis University of Oslo sltarifa@ifi.uio.no 29.11.2013 S. Lizeth Tapia Tarifa Outline Motivation 1 Motivation
More informationFast Software-managed Code Decompression
Fast Software-managed Code Decompression Charles Lefurgy and Trevor Mudge Advanced Computer Architecture Laboratory Electrical Engineering and Computer Science Dept. The University of Michigan, Ann Arbor
More informationSmart-Walk: An Intelligent Physiological Monitoring System for Smart Families
Smart-Walk: An Intelligent Physiological Monitoring System for Smart Families P. Sundaravadivel 1, S. P. Mohanty 2, E. Kougianos 3, V. P. Yanambaka 4, and M. K. Ganapathiraju 5 University of North Texas,
More informationPerformance of Fully Automated 3D Cracking Survey with Pixel Accuracy based on Deep Learning
Performance of Fully Automated 3D Cracking Survey with Pixel Accuracy based on Deep Learning Kelvin C.P. Wang Oklahoma State University and WayLink Systems Corp. 2017-10-19, Copenhagen, Denmark European
More informationPredicting NBA Shots
Predicting NBA Shots Brett Meehan Stanford University https://github.com/brettmeehan/cs229 Final Project bmeehan2@stanford.edu Abstract This paper examines the application of various machine learning algorithms
More informationAn Architecture for Combined Test Data Compression and Abort-on-Fail Test
An Architecture for Combined Test Data Compression and Abort-on-Fail Test Erik Larsson and Jon Persson Embedded Systems Laboratory Department of Computer and Information Science Linköpings Universitet,
More informationIntelligent Decision Making Framework for Ship Collision Avoidance based on COLREGs
Intelligent Decision Making Framework for Ship Collision Avoidance based on COLREGs Seminar Trondheim June 15th 2017 Nordic Institute of Navigation Norwegian Forum for Autonomous Ships SINTEF Ocean, Trondheim
More informationCBC2 performance with switched capacitor DC-DC converter. systems meeting, 12/2/14
CBC2 performance with switched capacitor DC-DC converter systems meeting, 12/2/14 1 reminder of results from CBC1 +2.5V GND 1n can power CBC from single +2.5V supply 1uF DC-DC diff. clock (CMOS) 1 MHz
More informationTR Electronic Pressure Regulator. User s Manual
TR Electronic Pressure Regulator Page 2 of 13 Table of Contents Warnings, Cautions & Notices... 3 Factory Default Setting... 4 Quick Start Procedure... 5 Configuration Tab... 8 Setup Tab... 9 Internal
More informationGOLOMB Compression Technique For FPGA Configuration
GOLOMB Compression Technique For FPGA Configuration P.Hema Assistant Professor,EEE Jay Shriram Group Of Institutions ABSTRACT Bit stream compression is important in reconfigurable system design since it
More informationThe Evolution of Transport Planning
The Evolution of Transport Planning On Proportionality and Uniqueness in Equilibrium Assignment Michael Florian Calin D. Morosan Background Several bush-based algorithms for computing equilibrium assignments
More informationSPD Pressure Sensor Families
DATASHEET SPD Pressure Sensor Families 1/7 Introduction to the principals of Smart Pressure Devices 1 Basic principles Pressure sensors are based on the principle of bending a membrane caused by the pressure
More informationFolding Reticulated Shell Structure Wind Pressure Coefficient Prediction Research based on RBF Neural Network
International Conference on Information Sciences, Machinery, Materials and Energy (ICISMME 2015) Folding Reticulated Shell Structure Wind Pressure Coefficient Prediction Research based on RBF Neural Network
More informationA Brief History of the Development of Artificial Neural Networks
A Brief History of the Development of Artificial Neural Networks Prof. Bernard Widrow Department of Electrical Engineering Stanford University Baidu July 18, 2018 Prof. Widrow @ Berkeley A Brief History
More informationBayesian Optimized Random Forest for Movement Classification with Smartphones
Bayesian Optimized Random Forest for Movement Classification with Smartphones 1 2 3 4 Anonymous Author(s) Affiliation Address email 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29
More informationFun Neural Net Demo Site. CS 188: Artificial Intelligence. N-Layer Neural Network. Multi-class Softmax Σ >0? Deep Learning II
Fun Neural Net Demo Site CS 188: Artificial Intelligence Demo-site: http://playground.tensorflow.org/ Deep Learning II Instructors: Pieter Abbeel & Anca Dragan --- University of California, Berkeley [These
More informationHigh-Speed Batting Using a Multi-Jointed Manipulator
High-Speed Batting Using a Multi-Jointed Manipulator Taku Senoo Akio Namiki Masatoshi Ishikawa Department of Information Physics and Computing, University of Tokyo 7-3- Hongo, Bunkyo-ku, Tokyo 3-8656,
More informationMass Flow Controller (MFC) for Gases
Mass Flow Controller (MFC) for Gases Type 8713 can be combined with... Direct flow measurement by MEMS- Technology for nominal flow rates from 1 ml N /min to 8 l N /min (N 2 ) High accuracy and repeatability
More informationSpecifications for Synchronized Sensor Pipe Condition Assessment (AS PROVIDED BY REDZONE ROBOTICS)
Specifications for Synchronized Sensor Pipe Condition Assessment (AS PROVIDED BY REDZONE ROBOTICS) A. Scope of Work The work covered by these specifications consists of furnishing all materials, labor,
More informationTAKING IT TO THE EXTREME The WAGO-I/O-SYSTEM for extreme Applications
TAKING IT TO THE EXTREME The WAGO-I/O-SYSTEM for extreme Applications extreme Temperature...from -40 C to +70 C (-40 F to +158 F) extreme Vibration...up to 5G Acceleration extreme Isolation...up to 5kV
More informationPL estimation acc. to EN ISO
PL estimation acc. to EN ISO 3849- Example calculation for an application MAC Safety / Armin Wenigenrath, January 2007 Select the suitable standard for your application Reminder: The standards and the
More informationLecture 39: Training Neural Networks (Cont d)
Lecture 39: Training Neural Networks (Cont d) CS 4670/5670 Sean Bell Strawberry Goblet Throne (Side Note for PA5) AlexNet: 1 vs 2 parts Caffe represents caffe like the above image, but computes as if it
More informationDiscussion: Illusions of Sparsity by Giorgio Primiceri
Discussion: Illusions of Sparsity by Giorgio Primiceri Pablo Guerron-Quintana Boston College June, 2018 Pablo Guerron-Quintana Discussion 1 / 17 Executive Summary Motivation: Is regular coke (dense model)
More informationDatasheet: K-30 ASCII Sensor
Datasheet: K-30 ASCII Sensor The K30 ASCII sensor is a low cost, infrared and maintenance free transmitter module intended to be built into different host devices that require CO2 monitoring data. The
More informationAdvanced Test Equipment Rentals ATEC (2832) OMS 600
Established 1981 Advanced Test Equipment Rentals www.atecorp.com 800-404-ATEC (2832) OMS 600 Continuous partial discharge monitoring system for power generators and electrical motors Condition monitoring
More informationOC508 Programmable mv Calibrator
OC508_GBM_201609 OC508 Programmable mv Calibrator Owner s Manual Orbit Controls AG Zürcherstrasse 137 CH8952 Schlieren - ZH orbitcontrols.ch OC508 Programmable mv Calibrator Selection of DIN Thermocouples
More informationMarch 5-8, 2017 Hilton Phoenix / Mesa Hotel Mesa, Arizona Archive Session 5
March 5-8, 2017 Hilton Phoenix / Mesa Hotel Mesa, Arizona Archive Session 5 2017 BiTS Workshop Image: tonda / istock Copyright Notice The presentation(s)/poster(s) in this publication comprise the Proceedings
More informationMulticore Real-Time Scheduling
Multicore Real-Time Scheduling Bjorn Andersson and Dionisio de Niz Software Engineering Institute Carnegie Mellon University Pittsburgh, PA 15213 Copyright 2015 Carnegie Mellon University This material
More informationLayout Design II. Lecture Fall 2003
Layout Design II Lecture 6 18-322 Fall 2003 Roadmap Today: Layout Verification & design in the large Next week: Transistor sizing Wires Homework 1: Due Today Homework 2: Out Today, Due Sept 18 Lab 2: This
More informationSontek RiverSurveyor Test Plan Prepared by David S. Mueller, OSW February 20, 2004
Sontek RiverSurveyor Test Plan Prepared by David S. Mueller, OSW February 20, 2004 INTRODUCTION Sontek/YSI has introduced new firmware and software for their RiverSurveyor product line. Firmware changes
More informationRATE CONTROL SYSTEM FOR SOUNDING ROCKETS
RATE CONTROL SYSTEM FOR SOUNDING ROCKETS Josef Ettl, Johann Pfänder German Aerospace Center (DLR), D-834 Wessling, Germany e-mail: josef.ettl@dlr.de johann.pfaender@dlr.de, 1. ABSTRACT DLR/Moraba has a
More informationStep Detection Algorithm For Accurate Distance Estimation Using Dynamic Step Length
Step Detection Algorithm For Accurate Distance Estimation Using Dynamic Step Length Ahmad Abadleh ahmad_a@mutah.edu.jo Eshraq Al-Hawari eshraqh@mutah.edu.jo Esra'a Alkafaween Esra_ok@mutah.edu.jo Hamad
More informationCT PET-2018 Part - B Phd COMPUTER APPLICATION Sample Question Paper
CT PET-2018 Part - B Phd COMPUTER APPLICATION Sample Question Paper Note: All Questions are compulsory. Each question carry one mark. 1. Error detection at the data link layer is achieved by? [A] Bit stuffing
More informationSpacecraft Simulation Tool. Debbie Clancy JHU/APL
FSW Workshop 2011 Using Flight Software in a Spacecraft Simulation Tool Debbie Clancy JHU/APL debbie.clancy@jhuapl.edu 443-778-7721 Agenda Overview of RBSP and FAST Technical Challenges Dropping FSW into
More informationDecompression Method For Massive Compressed Files In Mobile Rich Media Applications
2010 10th IEEE International Conference on Computer and Information Technology (CIT 2010) Decompression Method For Massive Compressed Files In Mobile Rich Media Applications Houchen Li, Zhijie Qiu, Lei
More informationPredicting Horse Racing Results with TensorFlow
Predicting Horse Racing Results with TensorFlow LYU 1703 LIU YIDE WANG ZUOYANG News CUHK Professor, Gu Mingao, wins 50 MILLIONS dividend using his sure-win statistical strategy. News AlphaGO defeats human
More informationFailure Modes, Effects and Diagnostic Analysis
Failure Modes, Effects and Diagnostic Analysis Project: Temperature transmitter PR5337 / PR6337 / PR7501 with 4..20 ma output Customer: PR electronics A/S Rønde Denmark Contract No.: PR electronics A/S
More informationThe HumiSys. RH Generator. Operation. Applications. Designed, built, and supported by InstruQuest Inc.
The HumiSys RH Generator Designed, built, and supported by InstruQuest Inc. Versatile Relative Humidity Generation and Multi-Sensor System The new HumiSys with single or dual RH probes capabilities is
More informationHW #5: Digital Logic and Flip Flops
HW #5: Digital Logic and Flip Flops This homework will walk through a specific digital design problem in all its glory that you will then implement in this weeks lab. 1 Write the Truth Table (10 pts) Consider
More informationAC : MEASUREMENT OF HYDROGEN IN HELIUM FLOW
AC 2010-2145: MEASUREMENT OF HYDROGEN IN HELIUM FLOW Randy Buchanan, University of Southern Mississippi Christopher Winstead, University of Southern Mississippi Anton Netchaev, University of Southern Mississippi
More informationSoftware Design of the Stiquito Micro Robot
Software Design of the Stiquito Micro Robot Andrew McClain and James M. Conrad University of North Carolina at Charlotte jmconrad@uncc.edu Abstract The Stiquito robot is a small, six legged robot that
More informationSCREENING OF TOPOGRAPHIC FACTOR ON WIND SPEED ESTIMATION WITH NEURAL NETWORK ANALYSIS
The Seventh Asia-Pacific Conference on Wind Engineering, November 8-12, 2009, Taipei, Taiwan SCREENING OF TOPOGRAPHIC FACTOR ON WIND SPEED ESTIMATION WITH NEURAL NETWORK ANALYSIS Fumiaki Nagao 1 Minoru
More informationNeuro-evolving Maintain-Station Behavior for Realistically Simulated Boats
Neuro-evolving Maintain-Station Behavior for Realistically Simulated Boats Nathan A. Penrod Evolutionary Computing System Lab Dept. of Computer Science and Engineering University of Nevada Reno, NV 89557
More informationTechnical Explanation
Technical Explanation for ESI Pressure Sensors Technical Explanation What is a pressure transmitter? A pressure transmitter (also called pressure transducer or pressure converter) is a component used to
More informationGait Analysis using Neural Networks
Gait Analysis using Neural Networks Sai krishna WDS 1 1Student, (B. Tech) Department of Electronics and communication, Vellore Institute of Technology, Chennai, India. -----------------------------------------------------------------------***--------------------------------------------------------------------
More informationSENSORS Pressure CALDARO
SENSORS Standards Specials THE MMI & MMF SPECIALIST The Caldaro Way From the very beginning Caldaro has managed to establish itself as a highly reputable OEM supplier of tailored controller and sensor
More informationSCIENTIFIC DATA SYSTEMS, INC. Depth Tension Line Speed Panel. DTLS Manual
SCIENTIFIC DATA SYSTEMS, INC. Depth Tension Line Speed Panel DTLS Manual This document contains proprietary information. Copyright 2015 Scientific Data Systems, Inc. All rights reserved. 1 Depth Tension
More informationMass Flow Meter (MFM) for Gases
873 Mass Flow Meter (MFM) for Gases Direct flow measurement by MEMS- Technology for nominal flow rates from 1 ml N /min to 8 l N /min (N 2 ) High accuracy Short response time Compact design and digital
More information3. REQUIRED EQUIPMENT
TEC? 700-700 Materiel Test Procedure 5-2-540 White Sands Missile Range I. OBJECTIVE U. S. ARMY TEST AND EVALUATION COMMAND COMMON ENGINEERING TEST PROCEDURE MISSILEBORNE GAS OPERATED POWER SUPPLY TESTS
More informationAN-140. Protege WX SALLIS Integration Application Note
AN-140 Protege WX SALLIS Integration Application Note The specifications and descriptions of products and services contained in this document were correct at the time of printing. Integrated Control Technology
More informationCombined neural network and numerical modeling for extreme storm surges
Leichtweiß-Institute for Hydraulic Engineering and Water Resources Department of Hydromechanics and Coastal Engineering Platzhalter für Bild, Bild auf Titelfolie hinter das Logo einsetzen Combined neural
More informationA Novel Decode-Aware Compression Technique for Improved Compression and Decompression
A Novel Decode-Aware Compression Technique for Improved Compression and Decompression J. Suresh Babu, K. Tirumala Rao & P. Srinivas Department Of Electronics & Comm. Engineering, Nimra College of Engineering
More informationReal-time identification using gait pattern analysis on a standalone wearable accelerometer
Real-time identification using gait pattern analysis on a standalone wearable accelerometer Guglielmo Cola, Marco Avvenuti and Alessio Vecchio Dipartimento di Ingegneria dell Informazione, University of
More information2. Determine how the mass transfer rate is affected by gas flow rate and liquid flow rate.
Goals for Gas Absorption Experiment: 1. Evaluate the performance of packed gas-liquid absorption tower. 2. Determine how the mass transfer rate is affected by gas flow rate and liquid flow rate. 3. Consider
More information5G trials in PyeongChang and at CEA-Leti. Presenter: Dr. Antonio CLEMENTE Wireless Communication Department, CEA-Leti, France
5G trials in PyeongChang and at CEA-Leti Presenter: Dr. Antonio CLEMENTE Wireless Communication Department, CEA-Leti, France Third annual ITU IMT-2020 Workshop & Demo Day, Gèneve July 18 2018 Outline Introduction
More informationA Single Cycle Processor. ECE4680 Computer Organization and Architecture. Designing a Pipeline Processor. Overview of a Multiple Cycle Implementation
ECE468 Computer Organization and Architecture Designing a Pipeline Processor A Single Cycle Processor Jump nstruction op Main nstruction : Fetch op 3 mm6 func 5 5 5 ctr 3 m Rw busw -bit
More informationknn & Naïve Bayes Hongning Wang
knn & Naïve Bayes Hongning Wang CS@UVa Today s lecture Instance-based classifiers k nearest neighbors Non-parametric learning algorithm Model-based classifiers Naïve Bayes classifier A generative model
More informationSafety Manual VEGAVIB series 60
Safety Manual VEGAVIB series 60 NAMUR Document ID: 32005 Contents Contents 1 Functional safety... 3 1.1 General information... 3 1.2 Planning... 4 1.3 Adjustment instructions... 6 1.4 Setup... 6 1.5 Reaction
More informationA Study on Algorithm for Compression and Decompression of Embedded Codes using Xilinx
A Study on Algorithm for Compression and Decompression of Embedded Codes using Xilinx M.Mohana Soundarya 1, Dr.S.Jayachitra 2 1 PG Scholar, Department of ECE, Vivekanandha College of Engineering for Women,
More informationWhat safety level can be reached when combining a contactor with a circuitbreaker for fail-safe switching?
FAQ 01/2015 What safety level can be reached when combining a contactor with a circuitbreaker for fail-safe switching? SIRIUS Safety Integrated http://support.automation.siemens.com/ww/view/en/40349715
More informationNaïve Bayes. Robot Image Credit: Viktoriya Sukhanova 123RF.com
Naïve Bayes These slides were assembled by Byron Boots, with only minor modifications from Eric Eaton s slides and grateful acknowledgement to the many others who made their course materials freely available
More informationDeZURIK. KGC Cast Knife Gate Valve. Safety Manual
KGC Cast Knife Gate Valve Safety Manual Manual D11036 August 29, 2014 Table of Contents 1 Introduction... 3 1.1 Terms... 3 1.2 Abbreviations... 4 1.3 Product Support... 4 1.4 Related Literature... 4 1.5
More informationExperimental Investigation of Dynamic Load Control Strategies using Active Microflaps on Wind Turbine Blades.
170 Experimental Investigation of Dynamic Load Control Strategies using Active Microflaps on Wind Turbine Blades. Authors & organisations F= First author P= Presenter Oliver Eisele oliver.eisele@fd.tu-berlin.de(1)
More informationintended velocity ( u k arm movements
Fig. A Complete Brain-Machine Interface B Human Subjects Closed-Loop Simulator ensemble action potentials (n k ) ensemble action potentials (n k ) primary motor cortex simulated primary motor cortex neuroprosthetic
More informationVLSI Design I; A. Milenkovic 1
Review: The Regenerative Property V i V o V i V o CPE/EE 47, CPE 57 VLSI esign I : Sequential Circuits epartment of Electrical and Computer Engineering University of labama in Huntsville leksandar Milenkovic
More informationIntroduction to Pattern Recognition
Introduction to Pattern Recognition Jason Corso SUNY at Buffalo 19 January 2011 J. Corso (SUNY at Buffalo) Introduction to Pattern Recognition 19 January 2011 1 / 32 Examples of Pattern Recognition in
More informationSensoric 4-20 ma Transmitter Board Operation Manual
Sensoric 4-20 ma Transmitter Board Operation Manual 1 Content Features Operation Manual Technical Data Mechanical Dimensions Remarks & Contact information 2 Features Soldered sensor cell (non replaceable)
More informationHeart Rate Prediction Based on Cycling Cadence Using Feedforward Neural Network
Heart Rate Prediction Based on Cycling Cadence Using Feedforward Neural Network Kusprasapta Mutijarsa School of Electrical Engineering and Information Technology Bandung Institute of Technology Bandung,
More informationFireHawk M7 Interface Module Software Instructions OPERATION AND INSTRUCTIONS
FireHawk M7 Interface Module Software Instructions OPERATION AND INSTRUCTIONS WARNING THE WARRANTIES MADE BY MSA WITH RESPECT TO THE PRODUCT ARE VOIDED IF THE PRODUCT IS NOT USED AND MAINTAINED IN ACCORDANCE
More informationCONTROL VALVE TESTING
The optimal functioning of the Control valve not only exists of sufficient body & seat tightness, but more important, the total "performance" of the valve and its controls! For an accurate and reliable
More informationIntroduction to Pattern Recognition
Introduction to Pattern Recognition Jason Corso SUNY at Buffalo 12 January 2009 J. Corso (SUNY at Buffalo) Introduction to Pattern Recognition 12 January 2009 1 / 28 Pattern Recognition By Example Example:
More informationDIY - PC - Interface for Suunto Cobra/Vyper/Mosquito
DIY - PC - Interface for Suunto Cobra/Vyper/Mosquito Summary This document is the distinct consequence of the Spyder/Stinger interface DIY. After having many e-mails concerning the application for the
More informationHot extractive multi-channel gas analyser for continuous process and emissions monitoring.
Product Data Sheet CODEL Continuous Emission Monitoring GCEM4100 Extractive Gas Analyser CO, CO2, NO, NO2, SO2, HCL, CH4 & H2O Hot extractive multi-channel gas analyser for continuous process and emissions
More informationOperating instructions Safety Rope Emergency Stop Switches ZB0052 / ZB0053 ZB0072 / ZB0073
Operating instructions Safety Rope Emergency Stop Switches UK ZB0052 / ZB0053 ZB0072 / ZB0073 7390878 / 02 03 / 2011 Contents 1 Safety instructions...3 2 Installation / set-up...4 2.1 Applications...4
More informationSession Objectives. At the end of the session, the participants should: Understand advantages of BFD implementation on S9700
BFD Features Session Objectives At the end of the session, the participants should: Understand advantages of BFD implementation on S9700 Understand when to use BFD on S9700 1 Contents BFD introduction
More informationAC : ACCURATE CRYOCHAMBER FOR A SMALL LABORATORY WITH SMALL BUDGET
AC 2007-1108: ACCURATE CRYOCHAMBER FOR A SMALL LABORATORY WITH SMALL BUDGET Matthew Braley, University of Idaho Paul Anderson, University of Idaho Tracey Windley, University of Idaho Kevin Buck, University
More informationReduction of Bitstream Transfer Time in FPGA
IOSR Journal of Electronics and Communication Engineering (IOSR-JECE) e-issn: 2278-2834,p- ISSN: 2278-8735.Volume 9, Issue 2, Ver. III (Mar - Apr. 2014), PP 82-86 Reduction of Bitstream Transfer Time in
More informationMitos Fluika Pressure and Vacuum Pumps Datasheet
Unit 1, Anglian Business Park, Orchard Road, Royston, Hertfordshire, SG8 5TW, UK T: +44 (0)1763 242491 F: +44 (0)1763 246125 E: sales@dolomite-microfluidics.com W: www.dolomite-microfluidics.com Dolomite
More informationInstrument pucks. Copyright MBARI Michael Risi SIAM design review November 17, 2003
Instrument pucks Michael Risi SIAM design review November 17, 2003 Instrument pucks Pucks and Plug-and-Work The MBARI puck prototype Puck software interface Pucks in practice (A Puck s Tale) Embedding
More informationNeural Nets Using Backpropagation. Chris Marriott Ryan Shirley CJ Baker Thomas Tannahill
Neural Nets Using Backpropagation Chris Marriott Ryan Shirley CJ Baker Thomas Tannahill Agenda Review of Neural Nets and Backpropagation Backpropagation: The Math Advantages and Disadvantages of Gradient
More informationADAPTING OMS SYSTEM TO DEAL WITH STORMS DATA IN REAL TIME
ADAPTING OMS SYSTEM TO DEAL WITH STORMS DATA IN REAL TIME André PAULO Cláudio MESQUITA Luís SANTOS Nuno SANTOS EDP Distribuição Portugal EDP Distribuição Portugal CGI Portugal CGI - Portugal andre.paulo@edp.pt
More informationPTP 800 SPLIT-MOUNT SOLUTION
TM PTP 800 SPLIT-MOUNT SOLUTION LICENSED ETHERNET MICROWAVE FOR MULTI-SERVICE NETWORKS Cambium Point-to-Point (PTP) 800 Licensed Ethernet Microwave Solutions can efficiently and affordably transport the
More information