Prof. Dr. Martin Brooke Bortecene Terlemez

Slides:



Advertisements
Similar presentations
ENEL111 Digital Electronics
Advertisements

VHDL - I 1 Digital Systems. 2 «The designer’s guide to VHDL» Peter J. Andersen Morgan Kaufman Publisher Bring laptop with installed Xilinx.
Neuromorphic Analog VLSI West Virginia University
Implementation Approaches with FPGAs Compile-time reconfiguration (CTR) CTR is a static implementation strategy where each application consists of one.
1 Power Management for High- speed Digital Systems Tao Zhao Electrical and Computing Engineering University of Idaho.
A reconfigurable system featuring dynamically extensible embedded microprocessor, FPGA, and customizable I/O Borgatti, M. Lertora, F. Foret, B. Cali, L.
Ultrasonic signal processing platform for nondestructive evaluation (NDE) Raymond Smith Advisors: Drs. In Soo Ahn, Yufeng Lu May 6, 2014.
University Of Vaasa Telecommunications Engineering Automation Seminar Signal Generator By Tibebu Sime 13 th December 2011.
A Dynamic Analog Concurrently-Processed Adaptive Chip Malcolm Stagg Grade 11.
Neural Cross Correlation For Radio Astronomy Chipo N Ngongoni Supervisor: Professor J Tapson Department of Electrical Engineering, University of Cape Town.
Neurmorphic Architectures Kenneth Rice and Tarek Taha Clemson University.
MICAS Department of Electrical Engineering (ESAT) AID–EMC: Low Emission Digital Circuit Design Status of the “Digital EMC project” Junfeng Zhou Wim Dehaene.
Prof. John Nestor ECE Department Lafayette College Easton, Pennsylvania ECE VLSI Circuit Design Lecture 25 - Subsystem.
Steven Koelmeyer BDS(hons)1 Reconfigurable Hardware for use in Ad Hoc Sensor Networks Supervisors Charles Greif Nandita Bhattacharjee.
Characterization Presentation Neural Network Implementation On FPGA Supervisor: Chen Koren Maria Nemets Maxim Zavodchik
6/11/2015 Adaptive Hardware Design for Digital Signal Processing Advisor: Dr. Thomas L. Stewart By: Prabjot Kaur Alex Tan.
Prof. John Nestor ECE Department Lafayette College Easton, Pennsylvania ECE VLSI Circuit Design Lecture 24 - Subsystem.
Chapter 10. Memory, CPLDs, and FPGAs
Moving NN Triggers to Level-1 at LHC Rates Triggering Problem in HEP Adopted neural solutions Specifications for Level 1 Triggering Hardware Implementation.
1 Lecture 16B Memories. 2 Memories in General Computers have mostly RAM ROM (or equivalent) needed to boot ROM is in same class as Programmable Logic.
February 4, 2002 John Wawrzynek
1 Lecture 16B Memories. 2 Memories in General RAM - the predominant memory ROM (or equivalent) needed to boot ROM is in same class as Programmable Logic.
1 EE 365 Adders Multipliers Read-Only Memories 2 Equality Comparators 1-bit comparator 4-bit comparator EQ_L.
Preliminary stuff Prof. Paul Hasler.
Engineering 1040: Mechanisms & Electric Circuits Fall 2011 Introduction to Embedded Systems.
1 A survey on Reconfigurable Computing for Signal Processing Applications Anne Pratoomtong Spring2002.
Using Programmable Logic to Accelerate DSP Functions 1 Using Programmable Logic to Accelerate DSP Functions “An Overview“ Greg Goslin Digital Signal Processing.
General FPGA Architecture Field Programmable Gate Array.
FPGA for Underwater Communication Pradyumna (Prad) Kadambi Mentor: Cody Youngbull April 13, 2015.
Challenges in Implementation of FPAA/FPGA Mixed-signal Technology
April 15, Synthesis of Signal Processing on FPGA Hongtao
MSE 2400 EaLiCaRA Spring 2015 Dr. Tom Way
1 Miodrag Bolic ARCHITECTURES FOR EFFICIENT IMPLEMENTATION OF PARTICLE FILTERS Department of Electrical and Computer Engineering Stony Brook University.
Lecture#14. Last Lecture Summary Memory Address, size What memory stores OS, Application programs, Data, Instructions Types of Memory Non Volatile and.
 Design model for a computer  Named after John von Neuman  Instructions that tell the computer what to do are stored in memory  Stored program Memory.
Floating Point vs. Fixed Point for FPGA 1. Applications Digital Signal Processing -Encoders/Decoders -Compression -Encryption Control -Automotive/Aerospace.
A comprehensive method for the evaluation of the sensitivity to SEUs of FPGA-based applications A comprehensive method for the evaluation of the sensitivity.
1 SERIAL PORT INTERFACE FOR MICROCONTROLLER EMBEDDED INTO INTEGRATED POWER METER Mr. Borisav Jovanović, Prof.dr Predrag Petković, Prof.dr. Milunka Damnjanović,
DLS Digital Controller Tony Dobbing Head of Power Supplies Group.
SYSTEM-ON-CHIP (SoC) AND USE OF VLSI CIRCUIT DESIGN TECHNOLOGY.
NIMIA October 2001, Crema, Italy - Vincenzo Piuri, University of Milan, Italy NEURAL NETWORKS FOR SENSORS AND MEASUREMENT SYSTEMS Part II Vincenzo.
1 Embedded Systems Computer Architecture. Embedded Systems2 Memory Hierarchy Registers Cache RAM Disk L2 Cache Speed (faster) Cost (cheaper per-byte)
VLSI & ECAD LAB Introduction.
Research on Reconfigurable Computing Using Impulse C Carmen Li Shen Mentor: Dr. Russell Duren February 1, 2008.
Neural Cross-Correlation For Radio Astronomy Chipo N Ngongoni Supervisor: Professor J Tapson Department of Electrical Engineering, University of Cape Town.
Neural Network Architectures Aydın Ulaş 02 December 2004
J. Christiansen, CERN - EP/MIC
“Politehnica” University of Timisoara Course No. 2: Static and Dynamic Configurable Systems (paper by Sanchez, Sipper, Haenni, Beuchat, Stauffer, Uribe)
An Optoelectronic Neural Network Packet Switch Scheduler K. J. Symington, A. J. Waddie, T. Yasue, M. R. Taghizadeh and J. F. Snowdon.
EE365 Adv. Digital Circuit Design Clarkson University Lecture #9 Math Units ROMs.
EE121 John Wakerly Lecture #6
Chapter 0 - reVieW Combinational Logic Circuit, Combinational Logic Circuit, Propagation Delay, Propagation Delay, Programmable Logic. Programmable Logic.
LOGIC OPTIMIZATION USING TECHNOLOGY INDEPENDENT MUX BASED ADDERS IN FPGA Project Guide: Smt. Latha Dept of E & C JSSATE, Bangalore. From: N GURURAJ M-Tech,
Kirchhoff Institute for Physics Johannes Schemmel Ruprecht-Karls-Universität Heidelberg 1 Accelerated Neuromorphic Hardware : Hybrid Plasticity - The Next.
Survey of multicore architectures Marko Bertogna Scuola Superiore S.Anna, ReTiS Lab, Pisa, Italy.
7/6/99 MITE1 Fully Parallel Learning Neural Network Chip for Real-time Control Students: (Dr. Jin Liu), Borte Terlemez Advisor: Dr. Martin Brooke.
Computer Architecture Lecture 26 Past and Future Ralph Grishman November 2015 NYU.
DEFENSE EXAMINATION GEORGIA TECH ECE P. 1 Fully Parallel Learning Neural Network Chip for Real-time Control Jin Liu Advisor: Dr. Martin Brooke Dissertation.
A Continuously Adapting Floating-Gate Node Jeff Dugger and Paul Hasler School of ECE Georgia Institute of Technology
Neural Networks. Background - Neural Networks can be : Biological - Biological models Artificial - Artificial models - Desire to produce artificial systems.
FPGA Field Programmable Gate Arrays Shiraz University of shiraz spring 2012.
FPGA Technology Overview Carl Lebsack * Some slides are from the “Programmable Logic” lecture slides by Dr. Morris Chang.
Digital Signal Processor HANYANG UNIVERSITY 학기 Digital Signal Processor 조 성 호 교수님 담당조교 : 임대현
Programmable Hardware: Hardware or Software?
Hardware Accelerator Test Bench for Error-Correcting Algorithms
Anne Pratoomtong ECE734, Spring2002
Prof. Carolina Ruiz Department of Computer Science
1.Introduction to Advanced Digital Design (14 marks)
♪ Embedded System Design: Synthesizing Music Using Programmable Logic
Prof. Carolina Ruiz Department of Computer Science
Presentation transcript:

Prof. Dr. Martin Brooke Bortecene Terlemez On-chip Learning Neural Network Hardware Implementation for Real-time Control Prof. Dr. Martin Brooke Bortecene Terlemez

Current Status Simulation Experiments Two frequency simulation Added noise simulation Experiments 1 second suppression Long runs

Simulation Setup Delay 1.5 ms Delay line error Unstable Combustion Model x u Software Simulation of Neural Network Chip

One Frequency Plant without Control

One Frequency Result f = 400Hz b = 

Two Frequency Results f = 400Hz 700Hz b =  The

10 % Added Noise Results f=400Hz z=0.005 b=1 Uncontrolled Engine Neural Network Controlled Engine f=400Hz z=0.005 b=1

Continuously Changing Plant Parameters (1 point/ second)

Continuously Changing Plant Parameters (50 points/ second)

Experimental Setup

Short run-time f = 400 Hz

Long run-time f = 400 Hz

Experimental Conclusions Suppression of Oscillation in less than few seconds. Continuous Adaptation.

Issues Competing technology status Controller Initialization General Purpose HW vs Dedicated HW Controller Initialization How to find optimum weights? How to set the weights?

Dedicated NN Hardware Serial Digital [1] Partially Parallel Digital [2] Fully Parallel Digital [3] Fully Parallel Analog [4]

References [1] Torsten Lehmann, Erik Bruun, and Casper Dietrich, “Mixed Analog/Digital Matrix-Vector Multiplier for Neural Network Synapses.” Analog Integrated Circuits and Signal Processing, 9, pp. 55-63, 1996. [2] Antonio J. Montalvo, Ronald S. Gyurcsik, and John J. Paulos, “An Analog VLSI Neural Network with On-Chip Perturbation Learning”, IEEE Journal of Solid-State Circuits, Vol. 32, No. 4, April 1997. [3] S. Neusser and B. Hofflinger, "Parallel Digital Neural Hardware for Controller Design", Mathematics and Computers in Simulation, Vol. 41, Pp. 149-160, 1996. [4] Maurizio Valle, Daniele D. Caviglia, and Ciacomo M. Bisio, “An Experimental Analog VLSI Neural Network with On-Chip Back-Propagation Learning”, Analog Integrated Circuits and Signal Processing, 9, pp. 231-245, 1996.

Time for One Forward Propagation Normalized speeds (Time: Number of Gate Delays)

(Area: Number of Transistors)

Today’s Technology - 0.35 mm CMOS Speed (ns) Chip Area (mm2)

Area and Time Requirement for 0.35-mm CMOS Process

Area and Time Estimation for 70-nm CMOS Process Speed (ns) Chip Area (mm2)

Area and Time Requirement for 70-nm CMOS Process

Controller Initialization How to find weights Simulation Is this good enough? Recorded training

Simulation Problem : Current chips are volatile Solution : FPGA

Recorded Simulation (current chip) Error Decrease Signal Random Sequence Error Decreases f = 400Hz z = 0.0 b = 0.1

Controller Initialization How to set weights Recorded simulation/training (current chips) permanent analog weight Digital weight storage (FPGA, custom)

Permanent Weight Storage EEPROM - FLASH Kah ng and S ze (?) FG De vi ce s and Circuits log 1989 1999 ET A N Br oo ke , et.al Shib ata/ Ohm i Ya STL AF G Ada pt iv e Ret ina ISD- Voic Re co rd er Digital Non-volitile Memories 1967 ...... .

Past EEPROM NN Permanent weight version of current chip

Permanent Analog Weight: Floating-Gate MOS (n-well) Regular CMOS Floating Gate MOS

RWC module with Floating-Gate MOS

Digital Weight Storage Custom digital chips Field Programmable Gate Arrays (FPGA)

Custom digital chips 13 bit programmable DAC 6-8 bits probably enough Expensive/slow to develop

Field Programmable Gate Arrays (FPGA) Reconfigurable Flexible Low-cost design cycle 1992: First ANN on FPGA 30 of XC3090 (8000 gates each) used Each neuron with 14 synapses:2 FPGA + 1 EPROM Today: very high density FPGAs with partial dynamic reconfiguration made possible ( >3 million gates)

RRANN Run-time Reconfigurable Artificial Neural Networks (RRANN) Time sharing the limited computing resource.

Conclusion FPGA technology ready Plan to attempt weight initialization Faster development Plan to adapt current test setups Plan to attempt weight initialization Recorded simulation/ training (current chips) Digital weights (FPGA)