G ENETIC A LGORITHMS FOR F AST M ATRIX M ULTIPLICATION András Joó Anikó Ekárt Juan Neirotti United Kingdom 14/07/2011 GECCO 2011 H UMIES AWARDS 2.

Slides:



Advertisements
Similar presentations
Divide-and-Conquer CIS 606 Spring 2010.
Advertisements

Science as a Process Chapter 1 Section 2.
11 Human Competitive Results of Evolutionary Computation Presenter: Mati Bot Course: Advance Seminar in Algorithms (Prof. Yefim Dinitz)
A simple example finding the maximum of a set S of n numbers.
Strassen's Matrix Multiplication Sibel KIRMIZIGÜL.
Divide-and-Conquer Matrix multiplication and Strassen’s algorithm Median Problem –In general finding the kth largest element of an unsorted list of numbers.
Lecture 10 Jianjun Hu Department of Computer Science and Engineering University of South Carolina CSCE350 Algorithms and Data Structure.
Genetic Algorithms Representation of Candidate Solutions GAs on primarily two types of representations: –Binary-Coded –Real-Coded Binary-Coded GAs must.
1 Evolvable Malware Sadia Noreen, Sahafq Murtaza, M. Zubair Shafiq, Muddassar Farooq National University of Computer and Emerging Sciences (FAST-NUCES)
Topological Interpretation of Crossover Alberto Moraglio & Riccardo Poli GECCO 2004.
1 Reduction between Transitive Closure & Boolean Matrix Multiplication Presented by Rotem Mairon.
Divide-and-Conquer Matrix multiplication and Strassen’s algorithm.
Genetic Algorithms GAs are one of the most powerful and applicable search methods available GA originally developed by John Holland (1975) Inspired by.
P OPULATION -B ASED I NCREMENTAL L EARNING : A Method for Integrating Genetic Search Based Function Optimization and Competitive Learning 吳昕澧 Date:2011/07/19.
Linear Simultaneous Equations
Taking a Square Root to Solve an Equation. Solve: In order to solve for x, you have to UNDO the squared first (i.e. square root) What are the number(s)
4.7 Identity and Inverse Matrices. What is an identity? In math the identity is the number you multiply by to have equivalent numbers. For multiplication.
LU Decomposition 1. Introduction Another way of solving a system of equations is by using a factorization technique for matrices called LU decomposition.
Automated Alphabet Reduction Method with Evolutionary Algorithms for Protein Structure Prediction Jaume Bacardit, Michael Stout, Jonathan D. Hirst, Kumara.
Numerical Computations in Linear Algebra. Mathematically posed problems that are to be solved, or whose solution is to be confirmed on a digital computer.
Basic Numerical methods and algorithms
Cristian Urs and Ben Riveira. Introduction The article we chose focuses on improving the performance of Genetic Algorithms by: Use of predictive models.
Copyright © 2011 Pearson Education, Inc. Publishing as Prentice Hall.
1 Science as a Process Chapter 1 Section 2. 2 Objectives  Explain how science is different from other forms of human endeavor.  Identify the steps that.
13.5 P ROPERTIES OF M ATRIX M ULTIPLICATION. W ARM -U P Use the following matrices to find 1 and 2 and 1. AB 2. BA.
1 Formal Verification of Candidate Solutions for Evolutionary Circuit Design (Entry 04) Zdeněk Vašíček and Lukáš Sekanina Faculty of Information Technology.
G ENETIC A LGORITHMS Ranga Rodrigo March 5,
Genetic algorithms and solid-state NMR pulse sequences Matthias Bechmann *, John Clark $, Angelika Sebald & * Department of Organic Chemistry, Johannes.
Introduction to Earth Science Section 2 Section 2: Science as a Process Preview Key Ideas Behavior of Natural Systems Scientific Methods Scientific Measurements.
R EPRESENTING H IGHER D IMENSIONAL A RRAYS INTO A G ENERALIZED T WO - DIMENSIONAL A RRAY Authors K.M. Azharul Hasan Md Abu Hanif Shaikh Dept. of Computer.
8.2 Operations With Matrices
Parallel Genetic Algorithms By Larry Hale and Trevor McCasland.
Solve each equation and check the solution. 08 September 2015 Warm-Up.
Sec 4.1 Matrices.
Genetic Programming as a Tool for novel Creation CS 621 Seminar Sri Raj Paul( ) Course Instructor Balamurali( ) Prof. Pushpak Bhattacharyya.
Assignment Questions?? Pg all, 23-26, 35, 46, 48 Handout Questions?
N- Queens Solution with Genetic Algorithm By Mohammad A. Ismael.
1 Complete this to a Pfaffian orientation (all internal faces have an odd number of clockwise arcs).
2.5 – Determinants and Multiplicative Inverses of Matrices.
CSE 421 Algorithms Richard Anderson Lecture 13 Recurrences, Part 2.
Warm-UP A = 7-310B = C =7-4Find:A 22 and C 31 97Find: the dimensions of each -88 Matrix Find: A + B and B – A and C + B.
1 SYSTEM OF LINEAR EQUATIONS BASE OF VECTOR SPACE.
Genetic Algorithms And other approaches for similar applications Optimization Techniques.
Genetic Algorithm(GA)
Review of Eigenvectors and Eigenvalues from CliffsNotes Online mining-the-Eigenvectors-of-a- Matrix.topicArticleId-20807,articleId-
Jaume Bacardit, Michael Stout, Jonathan D
Using GA’s to Solve Problems
An Evolutionary Approach
Evolvable Malware Sadia Noreen, Sahafq Murtaza, M. Zubair Shafiq, Muddassar Farooq National University of Computer and Emerging Sciences (FAST-NUCES) Next.
On Routine Evolution of Complex Cellular Automata
Genetic Algorithms GAs are one of the most powerful and applicable search methods available GA originally developed by John Holland (1975) Inspired by.
Zdeněk Vašíček and Lukáš Sekanina
Genetic Algorithms GAs are one of the most powerful and applicable search methods available GA originally developed by John Holland (1975) Inspired by.
Advanced Artificial Intelligence Evolutionary Search Algorithm
Quantum One.
CSCE 411 Design and Analysis of Algorithms
[ ] [ ] [ ] [ ] EXAMPLE 3 Scalar multiplication Simplify the product:
Topological Ordering Algorithm: Example
Richard Anderson Lecture 13 Recurrences and Divide and Conquer
EQ: How do I solve an equation in one variable?
Numerical Algorithms Quiz questions
Richard Anderson Lecture 12 Recurrences and Divide and Conquer
Choice A 90 students in a school have visited at least two other provinces. If this represents 24% of the students in the school, how many students are.
Richard Anderson Lecture 13 Recurrences, Part 2
CALCULATING EQUATION OF LEAST SQUARES REGRESSION LINE
Computer Simulation Techniques Generating Pseudo-Random Numbers
Topological Ordering Algorithm: Example
Topological Ordering Algorithm: Example
Matrix Multiplication Sec. 4.2
Topological Ordering Algorithm: Example
Presentation transcript:

G ENETIC A LGORITHMS FOR F AST M ATRIX M ULTIPLICATION András Joó Anikó Ekárt Juan Neirotti United Kingdom 14/07/2011 GECCO 2011 H UMIES AWARDS 2

T HE P ROBLEM : R ECURSIVE M ATRIX M ULTIPLICATION  Standard algorithm for multiplying two square matrices of size requires multiplications and d additions  Strassen’s algorithm reduces the number of required multiplications to if is a power of 2 (1969) 14/07/ GECCO 2011 H UMIES AWARDS

K NOWN L IMITS  For matrices of size at least 7 multiplications needed  For matrices of size at least 19 multiplications needed  Best known exact algorithm for size contains 23 multiplications 14/07/2011 GECCO 2011 H UMIES AWARDS 4

P RACTICAL S IGNIFICANCE  An exact algorithm using 22 multiplications on matrices of size would be an improvement on the best known algorithm for this size  An exact algorithm using 21 multiplications on matrices of size would be an overall improvement on how recursive matrix multiplication is currently performed on large matrices  As the search space has size 2.25e+180 for 21 multiplications and 8.71e+188 for 22 multiplications, respectively, it is highly unlikely that a human or a simple algorithm would discover a solution! 14/07/2011 GECCO 2011 H UMIES AWARDS 5

O UR SOLUTION : P ARALLEL GA  Parallel island model, with unidirectional ring topology and migration  Steady-state elitist GA  Continuous real-valued representation  Variety of crossover and mutation operators  Periodic explicit enforcing of diversity 14/07/2011 GECCO 2011 H UMIES AWARDS 6

GA R ESULTS On matrices of size  reproduced a solution with 23 multiplications  found an approximate solution of fitness for 22 multiplications 14/07/2011 GECCO 2011 H UMIES AWARDS 7

W HY H UMAN -C OMPETITIVE ?  In 1976, J. D. Laderman published his article “A noncommutative algorithm for multiplying matrices using 23 multiplications” in the Bulletin of the American Mathematical Society. Others published equivalent algorithms.  The theoretically proven lower bound is 19 multiplications, but no exact algorithm with less than 23 multiplications is known to date.  Our GA approach could reproduce matrix multiplication algorithms using 23 multiplications and also led to an approximate algorithm requiring 22 multiplications. 14/07/2011 GECCO 2011 H UMIES AWARDS 8

W HICH C RITERIA ?  B: The result is equal to or better than a result that was accepted as a new scientific result at the time when it was published in a peer-reviewed scientific journal.  D: The result is publishable in its own right as a new scientific result independent of the fact that the result was mechanically created.  F: The result is equal to or better than a result that was considered an achievement in its field at the time it was first discovered.  G: The result solves a problem of indisputable difficulty in its field. 14/07/2011 GECCO 2011 H UMIES AWARDS 9