dual entangled polynomial code three dimensional coding
play

Dual-Entangled Polynomial Code: Three-Dimensional Coding for - PowerPoint PPT Presentation

Jan 20, 2023 •427 likes •670 views

Dual-Entangled Polynomial Code: Three-Dimensional Coding for Distributed Matrix Multiplication Pedro Soto, Jun Li, Xiaodi Fan Florida International University Matrix Multiplication Matrix multiplication is a fundamental building block in

  1. Dual-Entangled Polynomial Code: Three-Dimensional Coding for Distributed Matrix Multiplication Pedro Soto, Jun Li, Xiaodi Fan Florida International University

  2. Matrix Multiplication ‣ Matrix multiplication is a fundamental building block in various machine learning algorithms. ‣ When the matrix comes from a large dataset, the multiplication will be split into smaller multiplications of submatrices on different nodes. � 2

  3. <latexit sha1_base64="sTfApjh8/Ig5dxgjSg4JiRvuvTc=">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</latexit> � � � Matrix Multiplication ‣ Matrix multiplication is a fundamental building block in various machine learning algorithms. ‣ When the matrix comes from a large dataset, the multiplication will be split into smaller multiplications of submatrices on different nodes. � 2

  4. <latexit sha1_base64="sTfApjh8/Ig5dxgjSg4JiRvuvTc=">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</latexit> � � � � �� � � � � � � <latexit sha1_base64="jvZ1pEC6o51a+h06RGOKnBfQifU=">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</latexit> Matrix Multiplication ‣ Matrix multiplication is a fundamental building block in various machine learning algorithms. ‣ When the matrix comes from a large dataset, the multiplication will be split into smaller multiplications of submatrices on different nodes. � 2

  5. <latexit sha1_base64="jvZ1pEC6o51a+h06RGOKnBfQifU=">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</latexit> <latexit sha1_base64="sTfApjh8/Ig5dxgjSg4JiRvuvTc=">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</latexit> � � � � � � �� � <latexit sha1_base64="t2WTPjNM0fu4hYh83muedb+5xhY=">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</latexit> <latexit sha1_base64="a10km9FQ/4m9DzwHTM1RECmc1mk=">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</latexit> � � � Matrix Multiplication ‣ Matrix multiplication is a fundamental building block in various machine learning algorithms. ‣ When the matrix comes from a large dataset, the multiplication will be split into smaller multiplications of submatrices on different nodes. worker � � � � worker � � � � � 2

Recommend

Entangled Polynomial Codes for Secure, Private, and Batch Distributed Matrix Multiplication:

Entangled Polynomial Codes for Secure, Private, and Batch Distributed Matrix Multiplication:

Entangled Polynomial Codes for Secure, Private, and Batch Distributed Matrix Multiplication: Breaking the &quot;Cubic&quot; Barrier QIAN YU (USC) joint work with Salman Avestimehr (USC) IEEE International Symposium on Information Theory (ISIT)

409 views • 21 slides
Coding for Everyone How your library can help anyone learn to code July 19, 2016 Kelly Smith

Coding for Everyone How your library can help anyone learn to code July 19, 2016 Kelly Smith

Coding for Everyone How your library can help anyone learn to code July 19, 2016 Kelly Smith Coding for Everyone Code Club Story Guiding Principles Coding for Pre-readers (3-7) Coding for Tweens (8-12) Coding for Teens

570 views • 30 slides
Chnotourg first formulation select each edge to either benthegraph ornotathegraphortuits dual

Chnotourg first formulation select each edge to either benthegraph ornotathegraphortuits dual

Chomutov Polynomial 50010.1 Towards a Symmetric Tutte Polynomial Chnotourg first formulation select each edge to either benthegraph ornotathegraphortuits dual Assign to each connected componentiealabelan the original graph org nthe dual graph

145 views • 12 slides
Pyrit: Polynomial Ring Transforms for Fast Erasure Coding Some parts of this work have been

Pyrit: Polynomial Ring Transforms for Fast Erasure Coding Some parts of this work have been

Pyrit: Polynomial Ring Transforms for Fast Erasure Coding Some parts of this work have been patented Jonathan Detchart and Jrme Lacan July 18, 2017 J. Detchart, J. Lacan NWCRG Erasure coding via polynomial ring transforms 1/ 14 Outline

316 views • 18 slides
Stochastic Computing by Stochastic Computing by a New Polynomial a New Polynomial Dimensional

Stochastic Computing by Stochastic Computing by a New Polynomial a New Polynomial Dimensional

th AIAA Conference, Palm Springs, CA, May 50 th 50 AIAA Conference, Palm Springs, CA, May 2009 2009 Stochastic Computing by Stochastic Computing by a New Polynomial a New Polynomial Dimensional Decom Dimensional Decomposition Method p

236 views • 20 slides
On Continuity of the Roots of a Parametric Zero Dimensional Multivariate Polynomial Ideal Yosuke

On Continuity of the Roots of a Parametric Zero Dimensional Multivariate Polynomial Ideal Yosuke

On Continuity of the Roots of a Parametric Zero Dimensional Multivariate Polynomial Ideal Yosuke Sato Ryoya Fukasaku Hiroshi Sekigawa Tokyo University of Science (Tokyo, Japan) 1 / 25 Motivation The roots of an unary polynomial are

1.11k views • 68 slides
Toward a Polynomial Model, Season III Polynomial Code Generation Paul Feautrier 1 Albert Cohen 2

Toward a Polynomial Model, Season III Polynomial Code Generation Paul Feautrier 1 Albert Cohen 2

Toward a Polynomial Model, Season III Polynomial Code Generation Paul Feautrier 1 Albert Cohen 2 Alain Darte 3 1 Ecole Normale Sup erieure de Lyon 2 Equipe INRIA Parkas 3 Xilinx and CNRS January 15, 2018 1 / 11 Polynomials Everywhere The

244 views • 11 slides
1 Two-Dimensional Parity Internet Checksum Use 1-dimensional parity Idea Add up all

1 Two-Dimensional Parity Internet Checksum Use 1-dimensional parity Idea Add up all

Error Coding Parity 1-bit error detection with parity Transmission process may introduce errors into a message. Add an extra bit to a code to ensure an even (odd) number of 1s Single bit errors versus burst errors Every

284 views • 4 slides
Bicrossed products whose dual has property (RD) but not polynomial growth Hua Wang Institut de

Bicrossed products whose dual has property (RD) but not polynomial growth Hua Wang Institut de

Bicrossed products whose dual has property (RD) but not polynomial growth Hua Wang Institut de Mathmatiques de Jussieu-Paris Rive Gauche Quantum groups and analysis workshop 59 August, 2019, Oslo Background and motivation General idea

598 views • 44 slides
Video Coding using Dual- Tree Wavelet Transform Beibei Wang 1 , Yao Wang 1 , Ivan Selesnick 1 ,

Video Coding using Dual- Tree Wavelet Transform Beibei Wang 1 , Yao Wang 1 , Ivan Selesnick 1 ,

Video Coding using Dual- Tree Wavelet Transform Beibei Wang 1 , Yao Wang 1 , Ivan Selesnick 1 , Anthony Vetro 2 1. Polytechnic University, Brooklyn, NY, 19020 2. MERL, Cambridge, MA 02139 Dual-tree DWT (DDWT) First proposed by Kingsbury,

714 views • 22 slides
? ? ? of maximal independent subsets of A . ? Examples of matroids: E is a family

? ? ? of maximal independent subsets of A . ? Examples of matroids: E is a family

Slide 1 Slide 3 Codes Finite geometry and permutation groups: An [ n , k ] code over GF ( q ) is a k -dimensional sub- space of GF ( q ) n . Its elements are called codewords . some polynomial links The weight wt ( v ) of v is the number of

421 views • 4 slides
Coding Lab: Why code? and getting situated Ari Anisfeld Summer 2020 1 / 24 2 / 24 Intro to

Coding Lab: Why code? and getting situated Ari Anisfeld Summer 2020 1 / 24 2 / 24 Intro to

Coding Lab: Why code? and getting situated Ari Anisfeld Summer 2020 1 / 24 2 / 24 Intro to coding lab Why are we here? What are we going to do? A quick introduction to R and R Studio and the tidyverse 3 / 24 Why coding? Many

269 views • 24 slides
Categories and Quantum Informatics Week 4: Dual objects Chris Heunen 1 / 37 Dual objects Idea:

Categories and Quantum Informatics Week 4: Dual objects Chris Heunen 1 / 37 Dual objects Idea:

Categories and Quantum Informatics Week 4: Dual objects Chris Heunen 1 / 37 Dual objects Idea: Quantum mechanically: maximally entangled states Graphically: bending wires 2 / 37 Dual objects Idea: Quantum mechanically: maximally

1.12k views • 75 slides
SD-DP: Sparse Dual of the Density Peaks Algorithm for Cluster Analysis of High-Dimensional Data

SD-DP: Sparse Dual of the Density Peaks Algorithm for Cluster Analysis of High-Dimensional Data

SD-DP: Sparse Dual of the Density Peaks Algorithm for Cluster Analysis of High-Dimensional Data November 5, 2018 Dimitris Floros 1 Tiancheng Liu 2 Nikos Pitsianis 12 Xiaobai Sun 2 1 Department of Electrical and Computer Engineering, Aristotle

685 views • 40 slides
Lecture 11 Vector Linear Network Coding Vector Linear Network Coding Outline Fundamentals for

Lecture 11 Vector Linear Network Coding Vector Linear Network Coding Outline Fundamentals for

Introduction to Network Coding Tuvi Etzion Lecture 11 Vector Linear Network Coding Vector Linear Network Coding Outline Fundamentals for vector network coding The combination network Vector network code vs. scalar network code 2 Multicast

657 views • 32 slides
Coding Theory Kaman Phamdo Mentor: Sean Ballentine December 9, 2015 What is a code? A code

Coding Theory Kaman Phamdo Mentor: Sean Ballentine December 9, 2015 What is a code? A code

Coding Theory Kaman Phamdo Mentor: Sean Ballentine December 9, 2015 What is a code? A code converts information into another representation Used for communication through a channel How computers communicate Encoding Decoding

440 views • 17 slides
Coding, Applied Control and Mechatronics Wi-Fi Code QR Code Scanner Accessing supports to date

Coding, Applied Control and Mechatronics Wi-Fi Code QR Code Scanner Accessing supports to date

Problem solving through Coding, Applied Control and Mechatronics Wi-Fi Code QR Code Scanner Accessing supports to date Webinar CPD Workshops To keep up-to-date: Join our mailing List This evening we will Appreciate how

665 views • 45 slides
Compression of Genetic Coding Sequences MohammadReza Ghodsi Genetic Code (Recap) The code

Compression of Genetic Coding Sequences MohammadReza Ghodsi Genetic Code (Recap) The code

Compression of Genetic Coding Sequences MohammadReza Ghodsi Genetic Code (Recap) The code defines a mapping between tri-nucleotide sequences called codons and amino acids. They begins with start codon (ATG), ends with a stop codon

421 views • 7 slides
Automatic Synthesis of Fast and Certified Code for Polynomial Evaluation through the example of

Automatic Synthesis of Fast and Certified Code for Polynomial Evaluation through the example of

ANR MetaLibm kick-off meeting Lyon, 22 January, 2014 Automatic Synthesis of Fast and Certified Code for Polynomial Evaluation through the example of the CGPE tool Guillaume Revy quipe-projet DALI, Univ. Perpignan Via Domitia LIRMM, CNRS:

570 views • 30 slides
Problem: Huffman Coding Def: binary character code = assignment of binary strings to characters

Problem: Huffman Coding Def: binary character code = assignment of binary strings to characters

Problem: Huffman Coding Def: binary character code = assignment of binary strings to characters e.g. ASCII code A = 01000001 fixed-length code B = 01000010 C = 01000011 How to decode: ? 01000001010000100100001101000001 Problem:

163 views • 13 slides
Problem: Huffman Coding Def: binary character code = assignment of binary strings to characters

Problem: Huffman Coding Def: binary character code = assignment of binary strings to characters

Problem: Huffman Coding Def: binary character code = assignment of binary strings to characters e.g. ASCII code A = 01000001 fixed-length code B = 01000010 C = 01000011 How to decode: ? 01000001010000100100001101000001 Problem:

348 views • 14 slides
Blaise Code Generator From implementing standards to coding automation ric Joyal September

Blaise Code Generator From implementing standards to coding automation ric Joyal September

Blaise Code Generator From implementing standards to coding automation ric Joyal September 24th, 2013 Overview Context Application Development Standards Code Builder Questionnaire Development Tool (QDT) Blaise Code

338 views • 31 slides
Lecture 15 with Shot Noise Chapter 10 Four- Dimensional Signal Con- stellations Dual-

Lecture 15 with Shot Noise Chapter 10 Four- Dimensional Signal Con- stellations Dual-

Lecture 15 Approximate Forms Based on Q Demodulation with Phase Noise Demodulation Lecture 15 with Shot Noise Chapter 10 Four- Dimensional Signal Con- stellations Dual- Polarization Modulation and De- modulation 1 ECE243b

554 views • 31 slides
BPSK with Block Coding Code rate is { } m b

BPSK with Block Coding Code rate is { } m b

BPSK with Block Coding Code rate is { } m b b m (1) ( ) , , R = k k &lt;1 n { } b Channel coding V f t r t cos(2 ) ( )

179 views • 13 slides

More recommend