1 optimization of vlsi circuit by genetic algorithms

Upload: idhawalvaja

Post on 02-Jun-2018

229 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/10/2019 1 Optimization of VLSI circuit by Genetic Algorithms

    1/18

    1

    Optimization of VLSI circuit byGenetic Algorithms

    Mirza Nazrul Alam

    GA Quick Overview

    What is genetic Algorithm :Genetic algorithm are computerizedsearch and optimization algorithmsbased on the mechanics of naturalgenetics and natural selection.genetic Algorithms are good at takinglarger, potentially huge, search spaceand navigating them looking for optimalsolutions which we might not find in a lifetime.

  • 8/10/2019 1 Optimization of VLSI circuit by Genetic Algorithms

    2/18

    2

    Idea of evolutionary computing was First Introduced byRechenberg in 1960 in his work Evolutionary strategies.

    Prof. Holland of University of Michigan, Ann Arborenvisaged the concept in the mid sixties and publishedhis seminal work ( Holland 1975).

    There after a number of students and others researchershave contributed to the development of this field.

    History:

    Three important aspects of using GA

    1 Definition of objective function2. Definition and implementationof genetic representation3. Definition and implementationof genetic operators.

  • 8/10/2019 1 Optimization of VLSI circuit by Genetic Algorithms

    3/18

    3

    Objective function

    The function f(x,y,z...) that has to beoptimized.Each parameter of the function has

    its own limit i.e., the lower and upperbound

    Genetic Representation

  • 8/10/2019 1 Optimization of VLSI circuit by Genetic Algorithms

    4/18

    4

    Search Space

    The space for all possible feasiblesolutions.Each solution is marked by its valueof fitness.

    Looking for solution means lookingfor extrema (either maximum orminimum ) in search space.

    Set of solutions in the search space. The initial populations are random

    numbers in the search space The populations are represented as

    chromosomes Binary ,octal or hexadecimal encoding

    are used to represent the chromosomes.

    Population

  • 8/10/2019 1 Optimization of VLSI circuit by Genetic Algorithms

    5/18

  • 8/10/2019 1 Optimization of VLSI circuit by Genetic Algorithms

    6/18

    6

    Basic Operators

    Reproduction or selectionCrossover

    mutation

    Reproduction

    Chromosomes are selected from thepopulation to crossover and produce

    offspring Obey the law of Darwin They best one should survive and

    create offspring.

  • 8/10/2019 1 Optimization of VLSI circuit by Genetic Algorithms

    7/18

    7

    Types of Reproduction

    Roulette-wheel selection Tournament Selection Rank selection

    Steady state selection

    Roulette-wheel Selection

  • 8/10/2019 1 Optimization of VLSI circuit by Genetic Algorithms

    8/18

    8

    Main idea: better individuals get higher chance Chances proportional to fitness Implementation: roulette wheel technique

    Assign to each individual a part of theroulette wheelSpin the wheel n times to select n

    individuals

    Roulette wheel

    fitness(A) = 3

    fitness(B) = 1

    fitness(C) = 2

    A C

    1/6 = 17%

    3/6 = 50%

    B2/6 = 33%

    Tournament competition among Nindividuals(N=2) are held atrandom.The highest fitness value is thewinner.Tournament is repeated untill themating pool for generating newoffspring is filled.

    Tournament Selection

  • 8/10/2019 1 Optimization of VLSI circuit by Genetic Algorithms

    9/18

    9

    Rank selection

    Roulette-wheel has problemwhen the fitness value differ verymuch.In rank selection the worst valuehas fitness 1, the next 2,......,and the best will have fitness N.

    1 23

    4

    5

    5

    1

    2

    3

    4

    5%2%

    8%

    10%

    75%

    33%

    7%

    13%

    20%

    27%

    Rank selection

    Roulette-wheel Rank selection

  • 8/10/2019 1 Optimization of VLSI circuit by Genetic Algorithms

    10/18

  • 8/10/2019 1 Optimization of VLSI circuit by Genetic Algorithms

    11/18

    11

    Single-point crossover

    Choose a random point on the two parentsSplit parents at this crossover pointCreate children by exchanging tailsP c typically in range (0.6, 0.9)

    n-point crossover

    Choose n random crossover pointsSplit along those pointsGlue parts, alternating between parents

    Generalisation of 1 point (still some positional bias)

  • 8/10/2019 1 Optimization of VLSI circuit by Genetic Algorithms

    12/18

    12

    Uniform crossover

    Assign 'heads' to one parent, 'tails' to the otherFlip a coin for each gene of the first childMake an inverse copy of the gene for the second childInheritance is independent of position

    operators: mutation

    Alter each gene independently with a probability p mp m is called the mutation rate Typically between 1/pop_size and 1/ chromosome_length

  • 8/10/2019 1 Optimization of VLSI circuit by Genetic Algorithms

    13/18

    13

    SGA technical summary table

    Emphasis on crossoverSpeciality

    All children replace parentsSurvivor selection

    Fitness-ProportionateParent selection

    Bitwise bit-flipping with fixedprobability

    Mutation

    N-point or uniformRecombination

    Binary stringsRepresentation

    SGA reproduction cycle

    1. Select parents for the mating pool(size of mating pool = population size)

    2. Shuffle the mating pool3. For each consecutive pair apply crossover with

    probability p c , otherwise copy parents4. For each offspring apply mutation (bit-flip with

    probability p m independently for each bit)5. Replace the whole population with the resulting

    offspring

  • 8/10/2019 1 Optimization of VLSI circuit by Genetic Algorithms

    14/18

  • 8/10/2019 1 Optimization of VLSI circuit by Genetic Algorithms

    15/18

    15

    X2 example: crossover

    X2 example: mutation

  • 8/10/2019 1 Optimization of VLSI circuit by Genetic Algorithms

    16/18

    16

    Applications of GA

    Design-VLSI circuit, aircraft design, communication network

    Robotics-Trajectory plannings Machine learning-designing neural network Control-pole balancing, misile evasion combinatorial optimization-travelling sales man, routing, bin packing , graph colouring.

    Optimization in VLSI circuit

    Designing VLSI circuits consistof two steps:

    1)topological design2) parameter determination .

  • 8/10/2019 1 Optimization of VLSI circuit by Genetic Algorithms

    17/18

    17

    Why GA in parameter optimizationof VLSI circuit?

    1)After designing a topology suitable for the circuit, thedesigner selects an appropriate value for each circuitelement from a circuit analysis.

    2) This step is difficult and time consuming becausedesigner must consider many factors simultaneously.

    3) Attempt has been made on how to simplify thisparameter design level using evolutionaryprogramming that is currently gaining significantattention in optimization technique.

    Concluding remarks

    It doesnt need a complex formulation for circuitanalysis. It is sufficient to select appropriateparts for optimization and to determine a

    suitable range of each parameter.It can be applied to the other circuits easily.When the specification of the circuit ischanged, it is only necessary to rewrite thecost function according to specification.

  • 8/10/2019 1 Optimization of VLSI circuit by Genetic Algorithms

    18/18

    Reference

    [1] Fogel, L.J., Owens, A.J., and Walsh, M.J., ArtificialIntelligence Through Simulated Evolution, John Wiley,Chichester, Uk, 1996.[2] Glover, F., Heuristics for Integer ProgrammingUsing Surrogate Constraints, Decision Sciences, Vol.8,No.1, pp.156-166, 1977.[3] Holland, J.H. Adaptation in Natural and ArtificialSystems, University of Michigan Press, Ann Arbor,1975.[4] Koza, J.R., Genetic Programming. MIT Press,Cambridge, MA, 1992.[5] Douglas A. Pucknell, Kamran Eshraghian, BasicVLSI Design, p-287