From Math to Efficient Hardware. James C. Hoe Department of ECE Carnegie Mellon University

Size: px
Start display at page:

Download "From Math to Efficient Hardware. James C. Hoe Department of ECE Carnegie Mellon University"

Transcription

1 FFT Compler: From Math to Effcet Hardware James C. Hoe Departmet of ECE Carege Mello Uversty jot wor wth Peter A. Mlder, Fraz Frachett, ad Marus Pueschel the SPIRAL project wth support from NSF ACR-3493, ITR/NGS-35687, DARPA NBCH-5,Itel The SPIRAL Project Hgh performace mplemetatos of lear DSP trasforms (DFT, DCT, DWT, flters, etc) are a mportat class of desg problems Had desg ad tug s trcy ad epesve - eeds both math ad mplemetato owledge - tme-cosumg ad tedous - eeds to repeat effort for every ew cotet SPIRAL research goal: A fleble push-butto desg geerato framewor that produces SW ad HW mplemetatos equal or better tha epert had desg Today I focus o FFT for HW Dataflow to Sythess Retrospectve, CSAIL/MIT, May 7, Slde

2 Why we ca do better tha had desg SPIRAL s oly focused o lear DSP trasforms These trasforms are hghly structured, hghly regular ad very well uderstood mathematcally Algorthmc mplemetatos of a trasform ca be eumerated followg a ow set of rules For a gve objectve fucto ad mappg target, a computer geerates a soluto at least as good as the best huma effort by tryg eough mplemetatos Dataflow to Sythess Retrospectve, CSAIL/MIT, May 7, Slde 3 Outle SPIRAL Formula Framewor SPIRAL for HW FFT cores Coclusos Dataflow to Sythess Retrospectve, CSAIL/MIT, May 7, Slde 4

3 Dataflow to Sythess Retrospectve, CSAIL/MIT, May 7, Slde 5 Lear Trasforms Lear trasform s matr-vector multplcato - computg by defto taes O(N ) operatos - the matr has structure E.g. dscrete Fourer trasform: y DFT N y y ẏj.. y N-... N-. N j e π.. N- j.. N- Dataflow to Sythess Retrospectve, CSAIL/MIT, May 7, Slde 6 Fast Algorthms a fast algorthm factors the matr to a sequece of structured, sparse matrces cheaper sparse multples O(N log(n)) operatos E.g. Cooley-Tuey Factorzato of DFT 4 Matr formula represetato ( ) ( ) m m m m m L DFT I D I DFT DFT

4 Fast Fourer Trasform Algorthms Recursvely factorze by the Cooley-Tuey rule utl oly leaf cases rema (e.g. DFT r for rad-r) DFT 8 Epoetal umber of alteratves Each ruletree correspods a dfferet algorthm All cost O(N log(n)) 8 8 ( DFT I 4 ) D ( I DFT4 ) L 8 4 ( DFT I ) D I ( DFT I ) D ( I DFT ) 4 DFT 8 DFT 4 DFT DFT DFT ( ( 4 ) 8 L L DFT 8 DFT DFT 4 DFT DFT Dataflow to Sythess Retrospectve, CSAIL/MIT, May 7, Slde 7 Formula to HW (Combatoal) Gve where s: - apply, the - s a permutato permute - apply, tmes parallel - s a dagoal scale B A A A Dataflow to Sythess Retrospectve, CSAIL/MIT, May 7, Slde 8

5 How about good HW? Matr formulas have a atural mappg to dataflow ad hece combatoal datapath However, real hardware desgs must ft a gve resource costrat sequetal datapath that reuse avalable HW - detfy repeated erels - statate erels uder resource costrats - schedule computato to reuse statated erels We wat to do the aalyss, mappg ad schedulg at formula level, wth hgh-level algorthm owledge Dataflow to Sythess Retrospectve, CSAIL/MIT, May 7, Slde 9 Regular Structure for HW Smple regular structure emboded Pease FFT Eample: Dataflow to Sythess Retrospectve, CSAIL/MIT, May 7, Slde

6 Pease DFT Eample: DFT 8 stage stage stage 3 (formula s appled from rght to left) Dataflow to Sythess Retrospectve, CSAIL/MIT, May 7, Slde Horzotal foldg p put bypass regster a basele datapath for DFT (sas bt-reverse) degree of freedom: vertcal parallelsm - parameter p Dataflow to Sythess Retrospectve, CSAIL/MIT, May 7, Slde

7 Vertcal (V-)foldg accordg to p cost latecy Fe-graed cotrol over cost/latecy tradeoff Dataflow to Sythess Retrospectve, CSAIL/MIT, May 7, Slde 3 DFTge [Mlder, et al., DAC 5, FPGA 6] fuctoalty cotrol area/latecy tradeoff resource model Dataflow to Sythess Retrospectve, CSAIL/MIT, May 7, Slde 4

8 Cost Performace Tradeoff latecy (usec) slces slces slces DFT 64 DFT 56 DFT 4 Dataflow to Sythess Retrospectve, CSAIL/MIT, May 7, Slde 5 Structures of Iterest for Parameterzed Cores Smple regular structure emboded formula Product: Horzotal foldg Tesor Product: parallelsm or streamg vertcal foldg Strde permutato: rewrg or buffer/delay Dataflow to Sythess Retrospectve, CSAIL/MIT, May 7, Slde 6

9 Applcablty to other trasforms? DFT rad DFT rad r -D DFT WHT [ L ( I DFT )] [ T ( I DFT ) L ] DP H DCT (type II) [ A L ] L L P R R / r / r [ T ( I r DFT r ) L r ] [( I r WHT r ) L r ] Dataflow to Sythess Retrospectve, CSAIL/MIT, May 7, Slde 7 Coclusos Mathematcal approach to DSP trasform mplemetato Ecapsulatg doma owledge a doma specfc tool for % automatc desg geerato ad optmzato Geeralzable to other lear DSP trasforms Tha you (Arvd) Dataflow to Sythess Retrospectve, CSAIL/MIT, May 7, Slde 8

SPIRAL DSP Transform Compiler:

SPIRAL DSP Transform Compiler: SPIRAL DSP Trasform Compiler: Applicatio Specific Hardware Sythesis Peter A. Milder (peter.milder@stoybroo.edu) Fraz Frachetti, James C. Hoe, ad Marus Pueschel Departmet of ECE Caregie Mello Uiversity

More information

Short Vector SIMD Code Generation for DSP Algorithms

Short Vector SIMD Code Generation for DSP Algorithms Short Vector SMD Code Geerato for DSP Algorthms Fraz Frachett Chrstoph Ueberhuber Appled ad Numercal Mathematcs Techcal Uversty of Vea Austra Markus Püschel José Moura Electrcal ad Computer Egeerg Carege

More information

Short Vector SIMD Code Generation for DSP Algorithms

Short Vector SIMD Code Generation for DSP Algorithms Short Vector SMD Code Geerato for DSP Algorthms Fraz Frachett Chrstoph Ueberhuber Appled ad Numercal Mathematcs Techcal Uversty of Vea Austra Markus Püschel José Moura Electrcal ad Computer Egeerg Carege

More information

Area and Power Efficient Modulo 2^n+1 Multiplier

Area and Power Efficient Modulo 2^n+1 Multiplier Iteratoal Joural of Moder Egeerg Research (IJMER) www.jmer.com Vol.3, Issue.3, May-Jue. 013 pp-137-1376 ISSN: 49-6645 Area ad Power Effcet Modulo ^+1 Multpler K. Ptambar Patra, 1 Saket Shrvastava, Sehlata

More information

Machine Learning: Algorithms and Applications

Machine Learning: Algorithms and Applications /03/ Mache Learg: Algorthms ad Applcatos Florao Z Free Uversty of Boze-Bolzao Faculty of Computer Scece Academc Year 0-0 Lecture 3: th March 0 Naïve Bayes classfer ( Problem defto A trag set X, where each

More information

A PROCEDURE FOR SOLVING INTEGER BILEVEL LINEAR PROGRAMMING PROBLEMS

A PROCEDURE FOR SOLVING INTEGER BILEVEL LINEAR PROGRAMMING PROBLEMS ISSN: 39-8753 Iteratoal Joural of Iovatve Research Scece, Egeerg ad Techology A ISO 397: 7 Certfed Orgazato) Vol. 3, Issue, Jauary 4 A PROCEDURE FOR SOLVING INTEGER BILEVEL LINEAR PROGRAMMING PROBLEMS

More information

LP: example of formulations

LP: example of formulations LP: eample of formulatos Three classcal decso problems OR: Trasportato problem Product-m problem Producto plag problem Operatos Research Massmo Paolucc DIBRIS Uversty of Geova Trasportato problem The decso

More information

Face Recognition using Supervised & Unsupervised Techniques

Face Recognition using Supervised & Unsupervised Techniques Natoal Uversty of Sgapore EE5907-Patter recogto-2 NAIONAL UNIVERSIY OF SINGAPORE EE5907 Patter Recogto Project Part-2 Face Recogto usg Supervsed & Usupervsed echques SUBMIED BY: SUDEN NAME: harapa Reddy

More information

A New Newton s Method with Diagonal Jacobian Approximation for Systems of Nonlinear Equations

A New Newton s Method with Diagonal Jacobian Approximation for Systems of Nonlinear Equations Joural of Mathematcs ad Statstcs 6 (3): 46-5, ISSN 549-3644 Scece Publcatos A New Newto s Method wth Dagoal Jacoba Appromato for Systems of Nolear Equatos M.Y. Wazr, W.J. Leog, M.A. Hassa ad M. Mos Departmet

More information

Transistor/Gate Sizing Optimization

Transistor/Gate Sizing Optimization Trasstor/Gate Szg Optmzato Gve: Logc etwork wth or wthout cell lbrary Fd: Optmal sze for each trasstor/gate to mmze area or power, both uder delay costrat Statc szg: based o tmg aalyss ad cosder all paths

More information

Fitting. We ve learned how to detect edges, corners, blobs. Now what? We would like to form a. compact representation of

Fitting. We ve learned how to detect edges, corners, blobs. Now what? We would like to form a. compact representation of Fttg Fttg We ve leared how to detect edges, corers, blobs. Now what? We would lke to form a hgher-level, h l more compact represetato of the features the mage b groupg multple features accordg to a smple

More information

Optimal Allocation of Complex Equipment System Maintainability

Optimal Allocation of Complex Equipment System Maintainability Optmal Allocato of Complex Equpmet System ataablty X Re Graduate School, Natoal Defese Uversty, Bejg, 100091, Cha edcal Protecto Laboratory, Naval edcal Research Isttute, Shagha, 200433, Cha Emal:rexs841013@163.com

More information

CS 2710 Foundations of AI Lecture 22. Machine learning. Machine Learning

CS 2710 Foundations of AI Lecture 22. Machine learning. Machine Learning CS 7 Foudatos of AI Lecture Mache learg Mlos Hauskrecht mlos@cs.ptt.edu 539 Seott Square Mache Learg The feld of mache learg studes the desg of computer programs (agets) capable of learg from past eperece

More information

COMBINATORIAL METHOD OF POLYNOMIAL EXPANSION OF SYMMETRIC BOOLEAN FUNCTIONS

COMBINATORIAL METHOD OF POLYNOMIAL EXPANSION OF SYMMETRIC BOOLEAN FUNCTIONS COMBINATORIAL MTHOD O POLYNOMIAL XPANSION O SYMMTRIC BOOLAN UNCTIONS Dala A. Gorodecky The Uted Isttute of Iformatcs Prolems of Natoal Academy of Sceces of Belarus, Msk,, Belarus, dala.gorodecky@gmal.com.

More information

Fuzzy Multi-objective Linear Programming Approach for Traveling Salesman Problem

Fuzzy Multi-objective Linear Programming Approach for Traveling Salesman Problem Fuzzy Mult-objectve Lear Programmg Approach for Travelg Salesma Problem Ama Rehmat Pujab Uversty College of Iformato Techology Uversty of the Pujab, Lahore, Pasta ama_mmal@yahoo.com Ha Saeed Pujab Uversty

More information

A Comparison of Univariate Smoothing Models: Application to Heart Rate Data Marcus Beal, Member, IEEE

A Comparison of Univariate Smoothing Models: Application to Heart Rate Data Marcus Beal, Member, IEEE A Comparso of Uvarate Smoothg Models: Applcato to Heart Rate Data Marcus Beal, Member, IEEE E-mal: bealm@pdx.edu Abstract There are a umber of uvarate smoothg models that ca be appled to a varety of olear

More information

ECE Digital Image Processing and Introduction to Computer Vision

ECE Digital Image Processing and Introduction to Computer Vision ECE59064 Dgtal Image Processg ad Itroducto to Computer Vso Depart. of ECE NC State Uverst Istructor: Tafu Matt Wu Sprg 07 Outle Recap Le Segmet Detecto Fttg Least square Total square Robust estmator Hough

More information

Airline Fleet Routing and Flight Scheduling under Market Competitions. Tang and Ming-Chei

Airline Fleet Routing and Flight Scheduling under Market Competitions. Tang and Ming-Chei Arle Fleet Routg ad Flght Schedulg uder Market Compettos Shagyao Ya, Ch-Hu Tag ad Mg-Che Lee Departmet of Cvl Egeerg, Natoal Cetral Uversty 3/12/2009 Itroducto Lterature revew The model Soluto method Numercal

More information

FEATURE SELECTION ON COMBINATIONS FOR EFFICIENT LEARNING FROM IMAGES. Rong Xiao, Lei Zhang, and Hong-Jiang Zhang

FEATURE SELECTION ON COMBINATIONS FOR EFFICIENT LEARNING FROM IMAGES. Rong Xiao, Lei Zhang, and Hong-Jiang Zhang FEATURE SELECTION ON COMBINATIONS FOR EFFICIENT LEARNING FROM IMAGES Rog Xao, Le Zhag, ad Hog-Jag Zhag Mcrosoft Research Asa, Bejg 100080, P.R. Cha {t-rxao, lezhag, hjzhag}@mcrosoft.com ABSTRACT Due to

More information

An Enhanced Local Covering Approach for Minimization of Multiple-Valued Input Binary-Valued Output Functions

An Enhanced Local Covering Approach for Minimization of Multiple-Valued Input Binary-Valued Output Functions Proceedgs of the 10th WSEAS Iteratoal Coferece o COMPUTERS, Voulagme, Athes, Greece, July 13-15, 2006 (pp63-68) A Ehaced Local Coverg Approach for Mmzato of Multple-Valued Iput Bary-Valued Output Fuctos

More information

Differentiated Service of Streaming Media Playback Technology

Differentiated Service of Streaming Media Playback Technology Iteratoal Coferece o Advaced Iformato ad Commucato Techology for Educato (ICAICTE 2013) Dfferetated Servce of Streamg Meda Playback Techology CHENG Z-ao 1 MENG Bo 1 WANG Da-hua 1 ZHAO Yue 1 1 Iformatzato

More information

Hardware/Software partitioning approach for embedded system design based on Genetic Algorithm

Hardware/Software partitioning approach for embedded system design based on Genetic Algorithm Iteratoal Joural of Scetfc Research & Egeerg Techology (IJSET ISSN: 2356-5608, Vol.3, ssue 3 Copyrght IPCO-205-pp.20-25 Hardware/Software parttog approach for embedded system desg based o Geetc Algorthm

More information

COMPARISON OF PARAMETERIZATION METHODS USED FOR B-SPLINE CURVE INTERPOLATION

COMPARISON OF PARAMETERIZATION METHODS USED FOR B-SPLINE CURVE INTERPOLATION Europea Joural of Techc COMPARISON OF PARAMETERIZATION METHODS USED FOR B-SPLINE CURVE INTERPOLATION Sıtı ÖZTÜRK, Cegz BALTA, Melh KUNCAN 2* Kocael Üverstes, Mühedsl Faültes, Eletro ve Haberleşme Mühedslğ

More information

Biconnected Components

Biconnected Components Presetato for use wth the textbook, Algorthm Desg ad Applcatos, by M. T. Goodrch ad R. Tamassa, Wley, 2015 Bcoected Compoets SEA PVD ORD FCO SNA MIA 2015 Goodrch ad Tamassa Bcoectvty 1 Applcato: Networkg

More information

Designing a learning system

Designing a learning system CS 75 Mache Leafrg Lecture 3 Desgg a learg system Mlos Hauskrecht mlos@cs.ptt.edu 539 Seott Square, x4-8845 people.cs.ptt.edu/~mlos/courses/cs75/ Homeork assgmet Homeork assgmet ll be out today To parts:

More information

Bezier curves. 1. Defining a Bezier curve. A closed Bezier curve can simply be generated by closing its characteristic polygon

Bezier curves. 1. Defining a Bezier curve. A closed Bezier curve can simply be generated by closing its characteristic polygon Curve represetato Copyrght@, YZU Optmal Desg Laboratory. All rghts reserved. Last updated: Yeh-Lag Hsu (--). Note: Ths s the course materal for ME55 Geometrc modelg ad computer graphcs, Yua Ze Uversty.

More information

Bicubic G 1 interpolation of arbitrary quad meshes using a 4-split

Bicubic G 1 interpolation of arbitrary quad meshes using a 4-split Bcubc G terpolato of arbtrary quad meshes usg a 4-splt Stefae Hahma, Georges-Perre Boeau, Baptste Caramaux Laboratore Jea Kutzma, Greoble Uersty, Frace Abstract We preset a pecewse b-cubc parametrc G sple

More information

Journal of Chemical and Pharmaceutical Research, 2015, 7(3): Research Article

Journal of Chemical and Pharmaceutical Research, 2015, 7(3): Research Article Avalable ole www.ocpr.com Joural of Chemcal ad Pharmaceutcal Research, 2015, 73):476-481 Research Artcle ISSN : 0975-7384 CODENUSA) : JCPRC5 Research o cocept smlarty calculato method based o sematc grd

More information

Eight Solved and Eight Open Problems in Elementary Geometry

Eight Solved and Eight Open Problems in Elementary Geometry Eght Solved ad Eght Ope Problems Elemetary Geometry Floret Smaradache Math & Scece Departmet Uversty of New Mexco, Gallup, US I ths paper we revew eght prevous proposed ad solved problems of elemetary

More information

Parallel Ant Colony for Nonlinear Function Optimization with Graphics Hardware Acceleration

Parallel Ant Colony for Nonlinear Function Optimization with Graphics Hardware Acceleration Proceedgs of the 009 IEEE Iteratoal Coferece o Systems Ma ad Cyberetcs Sa Atoo TX USA - October 009 Parallel At Coloy for Nolear Fucto Optmzato wth Graphcs Hardware Accelerato Wehag Zhu Departmet of Idustral

More information

Software reliability is defined as the probability of failure

Software reliability is defined as the probability of failure Evolutoary Regresso Predcto for Software Cumulatve Falure Modelg: a comparatve study M. Beaddy, M. Wakrm & S. Aljahdal 2 : Dept. of Math. & Ifo. Equpe MMS, Ib Zohr Uversty Morocco. beaddym@yahoo.fr 2:

More information

Prof. Feng Liu. Winter /24/2019

Prof. Feng Liu. Winter /24/2019 Prof. Feg Lu Wter 209 http://www.cs.pd.edu/~flu/courses/cs40/ 0/24/209 Last Tme Feature detecto 2 Toda Feature matchg Fttg The followg sldes are largel from Prof. S. Lazebk 3 Wh etract features? Motvato:

More information

Parallelism in Spiral

Parallelism in Spiral Parallelism in Spiral Franz Franchetti and the Spiral team (only part shown) Electrical and Computer Engineering Carnegie Mellon University Joint work with Yevgen Voronenko Markus Püschel This work was

More information

Optimization of Light Switching Pattern on Large Scale using Genetic Algorithm

Optimization of Light Switching Pattern on Large Scale using Genetic Algorithm Optmzato of Lght Swtchg Patter o Large Scale usg Geetc Algorthm Pryaka Sambyal, Pawaesh Abrol 2, Parvee Lehaa 3,2 Departmet of Computer Scece & IT 3 Departmet of Electrocs Uversty of Jammu, Jammu, J&K,

More information

Mesh Connectivity Compression for Progressive-to-Lossless Transmission

Mesh Connectivity Compression for Progressive-to-Lossless Transmission Mesh Coectvty Compresso for Progressve-to-Lossless Trasmsso Pegwe Hao, Yaup Paer ad Ala Pearma ISSN 1470-5559 RR-05-05 Jue 005 Departmet of Computer Scece Mesh Coectvty Compresso for Progressve-to-Lossless

More information

Joint Runtime / Energy Optimization and Hardware / Software Partitioning of Linear Transforms

Joint Runtime / Energy Optimization and Hardware / Software Partitioning of Linear Transforms Joint Runtime / Energy Optimization and Hardware / Software Partitioning of Linear Transforms Paolo D Alberto, Franz Franchetti, Peter A. Milder, Aliaksei Sandryhaila, James C. Hoe, José M. F. Moura, Markus

More information

Improved MOPSO Algorithm Based on Map-Reduce Model in Cloud Resource Scheduling

Improved MOPSO Algorithm Based on Map-Reduce Model in Cloud Resource Scheduling Improved MOPSO Algorthm Based o Map-Reduce Model Cloud Resource Schedulg Heg-We ZHANG, Ka NIU *, J-Dog WANG, Na WANG Zhegzhou Isttute of Iformato Scece ad Techology, Zhegzhou 45000, Cha State Key Laboratory

More information

1-D matrix method. U 4 transmitted. incident U 2. reflected U 1 U 5 U 3 L 2 L 3 L 4. EE 439 matrix method 1

1-D matrix method. U 4 transmitted. incident U 2. reflected U 1 U 5 U 3 L 2 L 3 L 4. EE 439 matrix method 1 -D matrx method We ca expad the smple plae-wave scatterg for -D examples that we ve see to a more versatle matrx approach that ca be used to hadle may terestg -D problems. The basc dea s that we ca break

More information

Signal-Path Driven Partition and Placement for Analog Circuit

Signal-Path Driven Partition and Placement for Analog Circuit Sgal-Path Drve Partto ad Placemet for Aalog Crcut D Log Xalog Hog Sheq Dog EDA Lab Departmet of Computer Scece ad Techology Tsghua Uversty Beg 100084 Cha logd02@malstsghuaeduc; hxl-dcs@maltsghuaeduc; dogsq@maltsghuaeduc

More information

Weighting Cache Replace Algorithm for Storage System

Weighting Cache Replace Algorithm for Storage System Weghtg Cache Replace Algorthm for Storage System Yhu Luo 2 Chagsheg Xe 2 Chegfeg Zhag 2 School of mathematcs ad Computer Scece, Hube Uversty, Wuha 430062, P.R. Cha 2 Natoal Storage System Laboratory, School

More information

Speeding- Up Fractal Image Compression Using Entropy Technique

Speeding- Up Fractal Image Compression Using Entropy Technique Avalable Ole at www.jcsmc.com Iteratoal Joural of Computer Scece ad Moble Computg A Mothly Joural of Computer Scece ad Iformato Techology ISSN 2320 088X IMPACT FACTOR: 5.258 IJCSMC, Vol. 5, Issue. 4, Aprl

More information

A Framework for Block-Based Timing Sensitivity Analysis

A Framework for Block-Based Timing Sensitivity Analysis 39.3 Framework for Block-Based Tmg Sestvty alyss Sajay V. Kumar Chadramoul V. Kashyap Sach S. Sapatekar Uversty of Mesota Itel Corporato Uversty of Mesota Meapols MN 55455 Hllsboro OR 973 Meapols MN 55455

More information

PERSPECTIVES OF THE USE OF GENETIC ALGORITHMS IN CRYPTANALYSIS

PERSPECTIVES OF THE USE OF GENETIC ALGORITHMS IN CRYPTANALYSIS PERSPECTIVES OF THE USE OF GENETIC ALGORITHMS IN CRYPTANALYSIS Lal Besela Sokhum State Uversty, Poltkovskaa str., Tbls, Georga Abstract Moder cryptosystems aalyss s a complex task, the soluto of whch s

More information

Self-intersection Avoidance for 3-D Triangular Mesh Model

Self-intersection Avoidance for 3-D Triangular Mesh Model Self-tersecto Avodace for 3-D Tragular Mesh Model Habtamu Masse Aycheh 1) ad M Ho Kyug ) 1) Departmet of Computer Egeerg, Ajou Uversty, Korea, ) Departmet of Dgtal Meda, Ajou Uversty, Korea, 1) hab01@ajou.ac.kr

More information

Cubic fuzzy H-ideals in BF-Algebras

Cubic fuzzy H-ideals in BF-Algebras OSR Joural of Mathematcs (OSR-JM) e-ssn: 78-578 p-ssn: 39-765X Volume ssue 5 Ver (Sep - Oct06) PP 9-96 wwwosrjouralsorg Cubc fuzzy H-deals F-lgebras Satyaarayaa Esraa Mohammed Waas ad U du Madhav 3 Departmet

More information

Using Linear-threshold Algorithms to Combine Multi-class Sub-experts

Using Linear-threshold Algorithms to Combine Multi-class Sub-experts Usg Lear-threshold Algorthms to Combe Mult-class Sub-experts Chrs Mesterharm MESTERHA@CS.RUTGERS.EDU Rutgers Computer Scece Departmet 110 Frelghuyse Road Pscataway, NJ 08854 USA Abstract We preset a ew

More information

Image Compression. CS 663, Ajit Rajwade

Image Compression. CS 663, Ajit Rajwade Image Compresso CS 663, Ajt Rajwade Image Compresso Process of covertg a mage fle to aother mage fle that occupes less storage space, wthout sacrfcg ts vsual cotet Useful for savg storage space, ad trasmsso

More information

Computation of DFT. W has a specific structure and because

Computation of DFT. W has a specific structure and because DSP 7 Computatio of DFT CTU EE Computatio of DFT Efficiet algoithms fo computig DFT Fast Fouie Tasfom. a Compute oly a few poits out of all poits b Compute all poits hat ae the efficiecy citeia? umbe of

More information

Clustering documents with vector space model using n-grams

Clustering documents with vector space model using n-grams Clusterg documets wth vector space model usg -grams Klas Skogmar, d97ksk@efd.lth.se Joha Olsso, d97jo@efd.lth.se Lud Isttute of Techology Supervsed by: Perre Nugues, Perre.Nugues@cs.lth.se Abstract Ths

More information

On a Sufficient and Necessary Condition for Graph Coloring

On a Sufficient and Necessary Condition for Graph Coloring Ope Joural of Dscrete Matheatcs, 04, 4, -5 Publshed Ole Jauary 04 (http://wwwscrporg/joural/ojd) http://dxdoorg/0436/ojd04400 O a Suffcet ad Necessary Codto for raph Colorg Maodog Ye Departet of Matheatcs,

More information

Vertex Odd Divisor Cordial Labeling of Graphs

Vertex Odd Divisor Cordial Labeling of Graphs IJISET - Iteratoal Joural of Iovatve Scece, Egeerg & Techology, Vol. Issue 0, October 0. www.jset.com Vertex Odd Dvsor Cordal Labelg of Graphs ISSN 48 68 A. Muthaya ad P. Pugaleth Assstat Professor, P.G.

More information

Workflow- Based Shape Optimization of Airfoils and Blades using Chained Bezier Curves

Workflow- Based Shape Optimization of Airfoils and Blades using Chained Bezier Curves Workflow- Based Shape Optmzato of Arfols ad Blades usg Chaed Bezer Curves Igor Pehec, Damr Vuča, Želja Loza Faculty of Electrcal Egeerg, Mechacal Egeerg ad Naval Archtecture FESB, Uversty of Splt, Croata

More information

Generating Parallel Transforms Using Spiral

Generating Parallel Transforms Using Spiral Generating Parallel Transforms Using Spiral Franz Franchetti Yevgen Voronenko Markus Püschel Part of the Spiral Team Electrical and Computer Engineering Carnegie Mellon University Sponsors: DARPA DESA

More information

Adaptive Clustering Algorithm for Mining Subspace Clusters in High-Dimensional Data Stream *

Adaptive Clustering Algorithm for Mining Subspace Clusters in High-Dimensional Data Stream * ISSN 673-948 CODEN JKYTA8 E-mal: fcst@vp.63.com Joural of Froters of Computer Scece ad Techology http://www.ceaj.org 673-948/200/04(09)-0859-06 Tel: +86-0-566056 DOI: 0.3778/j.ss.673-948.200.09.009 *,2,

More information

Some Results on Vertex Equitable Labeling

Some Results on Vertex Equitable Labeling Ope Joural of Dscrete Mathematcs, 0,, 5-57 http://dxdoorg/0436/odm0009 Publshed Ole Aprl 0 (http://wwwscrporg/oural/odm) Some Results o Vertex Equtable Labelg P Jeyath, A Maheswar Research Cetre, Departmet

More information

Data Structures and Algorithms(2)

Data Structures and Algorithms(2) Mg Zhag Data Structures ad Algorthms Data Structures ad Algorthms(2) Istructor: Mg Zhag Textbook Authors: Mg Zhag, Tegjao Wag ad Haya Zhao Hgher Educato Press, 2008.6 (the "Eleveth Fve-Year" atoal plag

More information

Integrated fault diagnosis method for down-hole working conditions of the beam pumping unit

Integrated fault diagnosis method for down-hole working conditions of the beam pumping unit Itegrated fault dagoss method for dow-hole workg codtos of the beam pumpg ut Ha Yg, L Ku College of Egeerg, Boha Uversty, Jzhou 203, Cha. E-mal: fegya326@alyu.com Abstract: The Dyamometer card s commoly

More information

Linearity, Slutsky symmetry, and a conjecture of Hicks. Abstract

Linearity, Slutsky symmetry, and a conjecture of Hicks. Abstract Learty Slutsy symmetry ad a coecture of Hcs Chrsta Weber Seattle Uversty Abstract Hcs (956) coectured that Slutsy symmetry should hold for dscrete as well as ftesmal rce chages f demad fuctos are globally

More information

Texture Feature Extraction using Slant-Hadamard Transform

Texture Feature Extraction using Slant-Hadamard Transform World Academy of Scece, Egeerg ad Techology 7 6 Texture Feature Extracto usg Slat- Trasform M. J. Nassr, A. Vafae, ad A. Moadjem Abstract Radom ad atural textures classfcato s stll oe of the bggest challeges

More information

Text Categorization Based on a Similarity Approach

Text Categorization Based on a Similarity Approach Text Categorzato Based o a Smlarty Approach Cha Yag Ju We School of Computer Scece & Egeerg, Uversty of Electroc Scece ad Techology of Cha, Chegdu 60054, P.R. Cha Abstract Text classfcato ca effcetly ehace

More information

Algorithms and Computation in Signal Processing

Algorithms and Computation in Signal Processing Algorithms and Computation in Signal Processing special topic course 8-799B spring 25 24 th and 25 th Lecture Apr. 7 and 2, 25 Instructor: Markus Pueschel TA: Srinivas Chellappa Research Projects Presentations

More information

Image Analysis. Segmentation by Fitting a Model. Christophoros Nikou We ve learned how to detect edges, corners, blobs. Now what?

Image Analysis. Segmentation by Fitting a Model. Christophoros Nikou We ve learned how to detect edges, corners, blobs. Now what? Image Aalyss Fttg Segmetato y Fttg a Model Chrstophoros Nkou ckou@cs.uo.gr Images take from: D. Forsyth ad J. Poce. Computer Vso: A Moder Approach, Pretce Hall, 003. Computer Vso course y Svetlaa Lazek,

More information

Enumerating XML Data for Dynamic Updating

Enumerating XML Data for Dynamic Updating Eumeratg XML Data for Dyamc Updatg Lau Ho Kt ad Vcet Ng Departmet of Computg, The Hog Kog Polytechc Uversty, Hug Hom, Kowloo, Hog Kog cstyg@comp.polyu.edu.h Abstract I ths paper, a ew mappg model, called

More information

Fuzzy ID3 Decision Tree Approach for Network Reliability Estimation

Fuzzy ID3 Decision Tree Approach for Network Reliability Estimation IJCSI Iteratoal Joural of Computer Scece Issues, Vol. 9, Issue 1, o 1, Jauary 2012 ISS (Ole): 1694-0814 www.ijcsi.org 446 Fuzzy ID3 Decso Tree Approach for etwor Relablty Estmato A. Ashaumar Sgh 1, Momtaz

More information

Summary of Curve Smoothing Technology Development

Summary of Curve Smoothing Technology Development RESEARCH ARTICLE Summary of Curve Smoothg Techology Developmet Wu Yze,ZhagXu,Jag Mgyag (College of Mechacal Egeerg, Shagha Uversty of Egeerg Scece, Shagha, Cha) Abstract: Wth the cotuous developmet of

More information

Signal Classification Method Based on Support Vector Machine and High-Order Cumulants

Signal Classification Method Based on Support Vector Machine and High-Order Cumulants Wreless Sesor Network,,, 48-5 do:.46/ws..7 Publshed Ole Jauary (http://www.scrp.org/joural/ws/). Sgal Classfcato Method Based o Support Vector Mache ad Hgh-Order Cumulats Abstract X ZHOU, Yg WU, B YANG

More information

SVM Classification Method Based Marginal Points of Representative Sample Sets

SVM Classification Method Based Marginal Points of Representative Sample Sets P P College P P College P Iteratoal Joural of Iformato Techology Vol. No. 9 005 SVM Classfcato Method Based Margal Pots of Represetatve Sample Sets Wecag ZhaoP P, Guagrog JP P, Ru NaP P, ad Che FegP of

More information

Towards Green Cloud Computing: Demand Allocation and Pricing Policies for. Cloud service brokerage.

Towards Green Cloud Computing: Demand Allocation and Pricing Policies for. Cloud service brokerage. 15 IEEE Iteratoal Coferece o Bg Data (Bg Data) Towards Gree Cloud Computg: Demad Allocato ad Prcg Polces for Cloud Servce Broerage Chex Qu, Hayg She ad Luhua Che Departmet of Electrcal ad Computer Egeerg

More information

Point Estimation-III: General Methods for Obtaining Estimators

Point Estimation-III: General Methods for Obtaining Estimators Pot Estmato-III: Geeral Methods for Obtag Estmators RECAP 0.-0.6 Data: Radom Sample from a Populato of terest o Real valued measuremets: o Assumpto (Hopefully Reasoable) o Model: Specfed Probablty Dstrbuto

More information

Using The ACO Algorithm in Image Segmentation for Optimal Thresholding 陳香伶財務金融系

Using The ACO Algorithm in Image Segmentation for Optimal Thresholding 陳香伶財務金融系 教專研 95P- Usg The ACO Algorthm Image Segmetato for Optmal Thresholdg Abstract Usg The ACO Algorthm Image Segmetato for Optmal Thresholdg 陳香伶財務金融系 Despte the fact that the problem of thresholdg has bee qute

More information

A Cordic-based Architecture for High Performance Decimal Calculations

A Cordic-based Architecture for High Performance Decimal Calculations A Cordc-ased Archtecture for Hgh Performace Decmal Calculatos Jose L. Sachez, Atoo Jmeo, Hgo Mora, Jeromo Mora, Fracsco Puol Computer Techology Departmet Uversty of Alcate Alcate, Spa Emal: sachez@dtc.ua.es,

More information

Mamdani Fuzzy Logic Controller with Mobile Agents for Matching

Mamdani Fuzzy Logic Controller with Mobile Agents for Matching RECENT DVNCES NEURL NETWORKS, FUZZY SYSTEMS & EVOLUTIONRY COMPUTING Mamda Fuzzy Logc Cotroller wth Moble gets for Matchg ION INCU, CLUDIU-IONUT POPIRLN Departmet of Computer Scece Uversty of Craova 3.

More information

Fuzzy Partition based Similarity Measure for Spectral Clustering. Xi an , China Abstract

Fuzzy Partition based Similarity Measure for Spectral Clustering. Xi an , China Abstract Iteratoal Joural of Sgal Processg, Image Processg ad Patter Recogto Vol.9, No., (6), pp.47-48 http://dx.do.org/.457/sp.6.9..39 Fuzzy Partto based Smlarty Measure for Spectral Clusterg Yfag Yag ad Yupg

More information

A Comparison of Heuristics for Scheduling Spatial Clusters to Reduce I/O Cost in Spatial Join Processing

A Comparison of Heuristics for Scheduling Spatial Clusters to Reduce I/O Cost in Spatial Join Processing Edth Cowa Uversty Research Ole ECU Publcatos Pre. 20 2006 A Comparso of Heurstcs for Schedulg Spatal Clusters to Reduce I/O Cost Spatal Jo Processg Jta Xao Edth Cowa Uversty 0.09/ICMLC.2006.258779 Ths

More information

MATHEMATICAL PROGRAMMING MODEL OF THE CRITICAL CHAIN METHOD

MATHEMATICAL PROGRAMMING MODEL OF THE CRITICAL CHAIN METHOD MATHEMATICAL PROGRAMMING MODEL OF THE CRITICAL CHAIN METHOD TOMÁŠ ŠUBRT, PAVLÍNA LANGROVÁ CUA, SLOVAKIA Abstract Curretly there s creasgly dcated that most of classcal project maagemet methods s ot sutable

More information

Assignment # 2. Farrukh Jabeen Algorithms 510 Assignment #2 Due Date: June 15, 2009.

Assignment # 2. Farrukh Jabeen Algorithms 510 Assignment #2 Due Date: June 15, 2009. Farrukh Jabeen Algorthms 51 Assgnment #2 Due Date: June 15, 29. Assgnment # 2 Chapter 3 Dscrete Fourer Transforms Implement the FFT for the DFT. Descrbed n sectons 3.1 and 3.2. Delverables: 1. Concse descrpton

More information

Flank Millable Surfaces Generated via Polynomial Composition

Flank Millable Surfaces Generated via Polynomial Composition 4 Flak Mllable Surfaces Geerate va Polyomal Composto Cheggag L, Stephe Ma 2 a Saeev Be 3 Uversty of Waterloo, cgl@egmal.uwaterloo.ca 2 Uversty of Waterloo, sma@uwaterloo.ca 3 Uversty of Waterloo, sbe@uwaterloo.ca

More information

Supplementary Information

Supplementary Information Supplemetary Iformato A Self-Trag Subspace Clusterg Algorthm uder Low-Rak Represetato for Cacer Classfcato o Gee Expresso Data Chu-Qu Xa 1, Ke Ha 1, Yog Q 1, Yag Zhag 2, ad Dog-Ju Yu 1,2, 1 School of Computer

More information

Impact of Mobility Prediction on the Temporal Stability of MANET Clustering Algorithms *

Impact of Mobility Prediction on the Temporal Stability of MANET Clustering Algorithms * Impact of Moblty Predcto o the Temporal Stablty of MANET Clusterg Algorthms * Aravdha Vekateswara, Vekatesh Saraga, Nataraa Gautam 1, Ra Acharya Departmet of Comp. Sc. & Egr. Pesylvaa State Uversty Uversty

More information

Exploring Wireless Sensor Network Configurations for Prolonging Network Lifetime

Exploring Wireless Sensor Network Configurations for Prolonging Network Lifetime 60 IJCSNS Iteratoal Joural of Computer Scece ad Network Securty, VOL7 No8, August 007 Explorg Wreless Sesor Network Cofguratos for Prologg Network Lfetme Zh Zha ad Yuhua Che, Departmet of Electrcal ad

More information

Computation of DFT. point DFT. W has a specific structure and because

Computation of DFT. point DFT. W has a specific structure and because DSP 5 Spig Computatio of DFT CTU EE Computatio of DFT Efficiet algoithms fo computig DFT Fast Fouie Tasfom. a Compute oly a few poits out of all poits b Compute all poits hat ae the efficiecy citeia? umbe

More information

Eight Solved and Eight Open Problems in Elementary Geometry

Eight Solved and Eight Open Problems in Elementary Geometry Eght Solved ad Eght Ope Problems Elemetary Geometry Floret Smaradache Math & Scece Departmet Uversty of New Mexco, Gallup, US I ths paper we revew eght prevous proposed ad solved problems of elemetary

More information

Computer Generation of IP Cores

Computer Generation of IP Cores A I n Computer Generation of IP Cores Peter Milder (ECE, Carnegie Mellon) James Hoe (ECE, Carnegie Mellon) Markus Püschel (CS, ETH Zürich) addfxp #(16, 1) add15282(.a(a69),.b(a70),.clk(clk),.q(t45)); addfxp

More information

Entropy-Based Collaborative Detection of DDOS Attacks on Community Networks

Entropy-Based Collaborative Detection of DDOS Attacks on Community Networks Sxth Aual IEEE Iteratoal Coferece o Pervasve Computg ad Commucatos Etropy-Based Collaboratve Detecto of DDOS Attacks o Commuty Networks Shu Yu ad Wale Zhou School of Egeerg ad Iformato Techology Deak Uversty,

More information

Classification Web Pages By Using User Web Navigation Matrix By Mementic Algorithm

Classification Web Pages By Using User Web Navigation Matrix By Mementic Algorithm Classfcato Web Pages By Usg User Web Navgato Matrx By Memetc Algorthm 1 Parvaeh roustae 2 Mehd sadegh zadeh 1 Studet of Computer Egeerg Software EgeergDepartmet of ComputerEgeerg, Bushehr brach,

More information

Optimal Multi-Image Processing Streaming Framework on Parallel Heterogeneous Systems

Optimal Multi-Image Processing Streaming Framework on Parallel Heterogeneous Systems Eurographcs Symposum o Parallel Graphcs ad Vsualzato (0) T. Kuhle, R. Pajarola, ad K. Zhou (Edtors) Optmal Mult-Image Processg Streamg Framework o Parallel Heterogeeous Systems Lh K. Ha, Jes Krüger, Joao

More information

A Quantitative Evaluation Method of Satellite Network Control Protocol based on Multi-attribute Utility Theory

A Quantitative Evaluation Method of Satellite Network Control Protocol based on Multi-attribute Utility Theory MATEC Web of Cofereces, A Quattate Ealuato Method of Satellte Network Cotrol Protocol based o Mult-attrbute Utlty Theory Ju Zhu ad Rum Lu 1 Nag Telecommucato Techology Research Isttute, Nag Jagsu 1, Cha

More information

Fingerprint Classification Based on Spectral Features

Fingerprint Classification Based on Spectral Features Fgerprt Classfcato Based o Spectral Features Hosse Pourghassem Tarbat Modares Uversty h_poorghasem@modares.ac.r Hassa Ghassema Tarbat Modares Uversty ghassem@modares.ac.r Abstract: Fgerprt s oe of the

More information

NEURO FUZZY MODELING OF CONTROL SYSTEMS

NEURO FUZZY MODELING OF CONTROL SYSTEMS NEURO FUZZY MODELING OF CONTROL SYSTEMS Efré Gorrosteta, Carlos Pedraza Cetro de Igeería y Desarrollo Idustral CIDESI, Av Pe de La Cuesta 70. Des. Sa Pablo. Querétaro, Qro, Méxco gorrosteta@teso.mx pedraza@cdes.mx

More information

NUMERICAL INTEGRATION BY GENETIC ALGORITHMS. Vladimir Morozenko, Irina Pleshkova

NUMERICAL INTEGRATION BY GENETIC ALGORITHMS. Vladimir Morozenko, Irina Pleshkova 5 Iteratoal Joural Iformato Theores ad Applcatos, Vol., Number 3, 3 NUMERICAL INTEGRATION BY GENETIC ALGORITHMS Vladmr Morozeko, Ira Pleshkova Abstract: It s show that geetc algorthms ca be used successfully

More information

PRIVATE set intersection (PSI) is a cryptographic protocol that

PRIVATE set intersection (PSI) is a cryptographic protocol that Effcet Delegated Prvate Set Itersecto o Outsourced Prvate Datasets Ayd Abad, Sotros Terzs, Roberto Metere, Chagyu Dog Abstract Prvate set tersecto (PSI) s a essetal cryptographc protocol that has may real

More information

Clustering Algorithm for High Dimensional Data Stream over Sliding Windows

Clustering Algorithm for High Dimensional Data Stream over Sliding Windows 2011 Iteratoal Jot Coferece of IEEE TrustCom-11/IEEE ICESS-11/FCST-11 Clusterg Algorthm for Hgh Dmesoal Data Stream over Sldg Wdows Weguo Lu, Ja OuYag School of Iformato Scece ad Egeerg, Cetral South Uversty,

More information

PEIECWISE CONSTANT LEVEL SET METHOD BASED FINITE ELEMENT ANALYSIS FOR STRUCTURAL TOPOLOGY OPTIMIZATION USING PHASE FIELD METHOD

PEIECWISE CONSTANT LEVEL SET METHOD BASED FINITE ELEMENT ANALYSIS FOR STRUCTURAL TOPOLOGY OPTIMIZATION USING PHASE FIELD METHOD INTERNATIONAL JOURNAL OF OPTIMIZATION IN CIVIL ENGINEERING It. J. Optm. Cvl Eg., 05; 5(4):389-407 PEIECWISE CONSTANT LEVEL SET METHOD BASED FINITE ELEMENT ANALYSIS FOR STRUCTURAL TOPOLOGY OPTIMIZATION

More information

A Genetic K-means Clustering Algorithm Applied to Gene Expression Data

A Genetic K-means Clustering Algorithm Applied to Gene Expression Data A Geetc K-meas Clusterg Algorthm Appled to Gee Expresso Data Fag-Xag Wu, W. J. Zhag, ad Athoy J. Kusal Dvso of Bomedcal Egeerg, Uversty of Sasatchewa, Sasatoo, S S7N 5A9, CANADA faw34@mal.usas.ca, zhagc@egr.usas.ca

More information

Machine Learning. CS 2750 Machine Learning. Administration. Lecture 1. Milos Hauskrecht 5329 Sennott Square, x4-8845

Machine Learning. CS 2750 Machine Learning. Administration. Lecture 1. Milos Hauskrecht 5329 Sennott Square, x4-8845 CS 75 Mache Learg Lecture Mache Learg Mlos Hauskrecht mlos@cs.ptt.edu 539 Seott Square, 5 people.cs.ptt.edu/~mlos/courses/cs75/ Admstrato Istructor: Prof. Mlos Hauskrecht mlos@cs.ptt.edu 539 Seott Square,

More information

Advanced Information Technology of Slot-Switching Network Schemes for on All-Optical Variable-Length Packet

Advanced Information Technology of Slot-Switching Network Schemes for on All-Optical Variable-Length Packet Joural of Computer Scece 6 (): -, 200 ISS 549-3636 200 Scece Publcatos Advaced Iformato Techology of Slot-Swtchg etwork Schemes for o All-Optcal Varable-Legth Packet Soug Yue Lew, 2 Edward Sek Kh Wog ad

More information

Mode-based temporal filtering for in-band wavelet video coding with spatial scalability

Mode-based temporal filtering for in-band wavelet video coding with spatial scalability Mode-based temporal flterg for -bad wavelet vdeo codg wth spatal scalablty ogdog Zhag a*, Jzheg Xu b, Feg Wu b, Weju Zhag a, ogka Xog a a Image Commucato Isttute, Shagha Jao Tog Uversty, Shagha b Mcrosoft

More information

A hybrid method using FAHP and TOPSIS for project selection Xuan Lia, Jiang Jiangb and Su Deng c

A hybrid method using FAHP and TOPSIS for project selection Xuan Lia, Jiang Jiangb and Su Deng c 5th Iteratoal Coferece o Computer Sceces ad Automato Egeerg (ICCSAE 205) A hybrd method usg FAHP ad TOPSIS for project selecto Xua La, Jag Jagb ad Su Deg c College of Iformato System ad Maagemet, Natoal

More information

A parallel implementation of the Durand-Kerner algorithm for polynomial root-finding on GPU

A parallel implementation of the Durand-Kerner algorithm for polynomial root-finding on GPU A parallel mplemetato of the Durad-Kerer algorthm for polyomal root-fdg o GPU Kaha Ghdouche, Raphaël Couturer, Abderrahmae Sder To cte ths verso: Kaha Ghdouche, Raphaël Couturer, Abderrahmae Sder. A parallel

More information

GUI Simulation Platform for RFID Indoor Tracking System

GUI Simulation Platform for RFID Indoor Tracking System Sesors & Trasducers 2014 by IFSA Publshg, S. L. http://www.sesorsportal.com GUI Smulato Platform for RFID Idoor Trackg System 1 Be-Be Mao, 2 JIN Xue-Bo School of Computer ad Iformato Egeerg, Bejg Techology

More information