Introduction to Genetic Algorithms

1
Introduction to Genetic Algorithms

• For CSE/ECE 848
• Introduction to Evolutionary Computation
• Prepared by Erik Goodman
• Professor, Electrical and Computer Engineering
• Michigan State University, and
• Co-Director, Genetic Algorithms Research and
  Applications Group (GARAGe)
• Based on and Accompanying Darrell Whitleys
  Genetic Algorithms Tutorial

2
Genetic Algorithms

• Are a method of search, often applied to
  optimization or learning
• Are stochastic but are not random search
• Use an evolutionary analogy, survival of
  fittest
• Not fast in some sense but sometimes more
  robust scale relatively well, so can be useful
• Have extensions including Genetic Programming
  (GP) (LISP-like function trees), learning
  classifier systems (evolving rules), linear GP
  (evolving ordinary programs), many others

3
The Canonical or Classical GA

• Maintains a set or population of strings at
  each stage
• Each string is called a chromosome, and encodes a
  candidate solution CLASSICALLY, encodes as a
  binary string (and now in almost any conceivable
  representation)

4
Criterion for Search

• Goodness (fitness) or optimality of a strings
  solution determines its FUTURE influence on
  search process -- survival of the fittest
• Solutions which are good are used to generate
  other, similar solutions which may also be good
  (even better)
• The POPULATION at any time stores ALL we have
  learned about the solution, at any point
• Robustness (efficiency in finding good solutions
  in difficult searches) is key to GA success

5
Classical GA The Representation

• 1011101010 a possible 10-bit string
  (CHROMOSOME) representing a possible solution
  to a problem
• Bits or subsets of bits might represent choice of
  some feature, for example. Lets represent
  choice of shipping container for some object
• bit position meaning
• 1-2 steel, aluminum, wood or cardboard
• 3-5 thickness (1mm-8mm)
• 6-7 fastening (tape, glue, rope, plastic
  wrap)
• 8 stuffing (paper or plastic peanuts)
• 9 corner reinforcement (yes, no)
• 10 handles (yes, no)

6
Terminology

• Each position (or each set of positions that
  encodes some feature) is called a LOCUS (plural
  LOCI)
• Each possible value at a locus is called an
  ALLELE
• We need a simulator, or evaluator program, that
  can tell us the (probable) outcome of shipping a
  given object in any particular type of container
• may be a COST (including losses from damage) (for
  example, maybe 1.4 means very low cost, 8.3 is
  very bad on a scale of 0-10.0), or
• may be a FITNESS, or a number that is larger if
  the result is BETTER

7
How Does a GA Operate?

• For ANY chromosome, must be able to determine a
  FITNESS (measure performance toward an objective)
• Objective may be maximized or minimized usually
  say fitness is to be maximized, and if objective
  is to be minimized, define fitness from it as
  something to maximize

8
GA OperatorsClassical Mutation

• Operates on ONE parent chromosome
• Produces an offspring with changes.
• Classically, toggles one bit in a binary
  representation
• So, for example 1101000110 could mutate
  to 1111000110
• Each bit has same probability of mutating

9
Classical Crossover

• Operates on two parent chromosomes
• Produces one or two children or offspring
• Classical crossover occurs at 1 or 2 points
• For example (1-point) (2-point)
• 1111111111 or 1111111111
• X 0000000000 0000000000
• 1110000000 1110000011
• and 0001111111 0001111100

10
Selection Operation

• Traditionally, parents are chosen to mate with
  probability proportional to their fitness
  proportional selection
• Traditionally, children replace their parents
• Many other variations now more commonly used
  (well come back to this)
• Overall principle survival of the fittest

11
Synergy the KEY

• Clearly, selection alone is no good
• Clearly, mutation alone is no good
• Clearly, crossover alone is no good
• Fortunately, using all three simultaneously is
  sometimes spectacular!

12
Canonical GA Differences from Other Search Methods

• Maintains a set or population of solutions at
  each stage (see blackboard)
• Classical or canonical GA always uses a
  crossover or recombination operator (domain is
  PAIRS of solutions (sometimes more))
• All we have learned to time t is represented by
  time ts POPULATION of solutions

13
Contrast with Other Search Methods

• indirect -- setting derivatives to 0
• direct -- hill climber (already described)
• enumerative -- already described
• random -- already described
• simulated annealing -- already described
• Tabu
• RSM -- fits approx. surf to set of pts, avoids
  full evaluations during local search

14
BEWARE of Claims about ANY Algorithms Asymptotic
Behavior Eventually is a LONG Time

• LOTS of methods can guarantee to find the best
  solution, probability 1, eventually
• Enumeration
• Random search (better without resampling)
• SA (properly configured)
• Any GA that avoids absorbing states in a Markov
  chain
• The POINT you cant afford to wait that long,
  if the problem is anything interesting!!!

15
When Might a GABe Any Good?

• Highly multimodal functions
• Discrete or discontinuous functions
• High-dimensionality functions, including many
  combinatorial ones
• Nonlinear dependencies on parameters
  (interactions among parameters) -- epistasis
  makes it hard for others
• Often used for approximating solutions to
  NP-complete combinatorial problems
• DONT USE if a hill-climber, etc., will work well

16
The Limits to Search

• No search method is best for all problems per
  the No Free Lunch Theorem
• Dont let anyone tell you a GA (or THEIR favorite
  method) is best for all problems!!!
• Needle-in-a-haystack is just hard, in practice
• Efficient search must be able to EXPLOIT
  correlations in the search space, or its no
  better than random search or enumeration
• Must balance with EXPLORATION, so dont just find
  nearest local optimum

17
Examples of Successful Real-World GA Application

• Antenna design
• Drug design
• Chemical classification
• Electronic circuits (Koza)
• Factory floor scheduling (Volvo, Deere, others)
• Turbine engine design (GE)
• Crashworthy car design (GM/Red Cedar)
• Protein folding

• Network design
• Control systems design
• Production parameter choice
• Satellite design
• Stock/commodity analysis/trading
• VLSI partitioning/ placement/routing
• Cell phone factory tuning
• Data Mining

18
EXAMPLE!!!Lets Design a Flywheel

• GOAL To store as much energy as possible (for a
  given diameter flywheel) without breaking apart
• On the chromosome, a number specifies the
  thickness (height) of the ring at each given
  radius
• Center hole for a bearing is fixed
• To evaluate simulate spinning it faster and
  faster until it breaks calculate how much energy
  is stored just before it breaks

19
Flywheel Example

• So if we use 8 rings, the chromosome might look
  like
• 6.3 3.7 2.5 3.5 5.6 4.5 3.6 4.1
• If we mutate HERE, we might get
• 6.3 3.7 4.1 3.5 5.6 4.5 3.6 4.1
• And that might look like (from the side)

20
Recombination (Crossover)

• If we recombine two designs, we might get
• 6.3 3.7 2.5 3.5 5.6 4.5 3.6 4.1
• x
• 3.6 5.1 3.2 4.3 4.4 6.2 2.3 3.4
• 3.6 5.1 3.2 3.5 5.6 4.5 3.6 4.1
• This new design might be BETTER or WORSE!

21
Typical GA Operation -- Overview
Initialize population at random
Evaluate fitness of new chromosomes
Good Enough?
Yes
Done
No
Select survivors (parents) based on fitness
Perform crossover and mutation on parents
22
A GA Evolves the Flywheel
One Choice of
Choice Material Materials
(side view)
23
Another Example NASA ST5 Quadrifilar Helical
AntennaGiven a Desired Pattern, Design the
Antenna

• Prior to Lohns evolution of a design, a contract
  had been awarded for designing the antenna.
• Result this quadrifilar helical antenna (QHA).

Radiator Under the ground plane matching and
phasing network
24
2nd Set of Evolved Antennas(Now Flying on 3
Satellites)
25
Genetic Algorithm -- Meaning?

• classical or canonical GA -- Holland (60s,
  book in 75) -- binary chromosome, population,
  selection, crossover (recombination), low rate
  mutation
• More general GA population, selection, (
  recombination) ( mutation) -- may be hybridized
  with LOTS of other stuff

26
Representation Terminology

• Classically, binary string individual or
  chromosome
• Whats on the chromosome is GENOTYPE
• What it means in the problem context is the
  PHENOTYPE (e.g., binary sequence may map to
  integers or reals, or order of execution, or
  inputs to a simulator, etc.)
• Genotype determines phenotype, but phenotype may
  look very different

27
Optimization Formulation

• Not all GAs used for optimization -- also
  learning, etc.
• Commonly formulated as given F(X1,Xn), find set
  of Xis (in a range) that extremize F, often also
  with additional constraint equations (equality or
  inequality) Gi(X1,Xn) lt Li, that must also be
  satisfied.
• Encoding obviously depends on problem being solved

28
Discretization Representation Meets Mutation!

• If problem is binary decisions, bit-flip mutation
  is fine
• BUT if using binary numbers to encode integers,
  as in 0,15 ? 0000, 1111, problem with Hamming
  cliffs
• One mutation can change 6 to 7 0110 ? 0111, BUT
• Need 4 bit-flips to change 7 to 8 0111 ? 1000
• Thats called a Hamming cliff
• May use Gray (or other distance-one) codes to
  improve properties of operators for example
  000, 001, 011, 010, 110, 111, 101, 100

29
Mutation Revisited

• On parameter encoded representations
• Binary ints
• Gray codes and bit-flips
• Or binary ints 0-mean, Gaussian changes, etc.
• Real-valued domain
• Can discretize to binary -- typically powers of 2
  with lower, upper limits, linear/exp/log scaling
• End result (classically) is a bit string
• BUT many now work with real-valued GAs,
  non-bit-flip (0-mean, Gaussian noise) mutation
  operators

30
Recombination or Crossover

• On parameter encoded representations
• 1-pt example
• 2-pt example
• uniform example
• Linkage loci nearby on chromosome, not usually
  disrupted by a given crossover operator (cf.
  1-pt, 2-pt, uniform re linkage)
• But use OTHER crossover operators for reordering
  problems (later)

31
Defining Objective/Fitness Functions

• Problem-specific, of course
• Many involve using a simulator
• Dont need to possess derivatives
• May be stochastic
• Need to evaluate thousands of times, so cant be
  TOO COSTLY

32
The What Function?

• In problem-domain form -- absolute or raw
  fitness, or evaluation or performance or
  objective function
• Relative (to population), possibly inverted
  and/or offset, scaled fitness usually called the
  fitness function. Fitness should be MAXIMIZED,
  whereas the objective function might need to be
  MAXIMIZED OR MINIMIZED.

33
Defining Objective/Fitness Functions

• Problem-specific, of course
• Many involve using a simulator
• Dont need to know (or even HAVE) derivatives
• May be stochastic
• Need to evaluate thousands of
• times, so cant be TOO
• COSTLY
• For real-world, evaluation
• time is typical bottleneck
• Example simple fitness
• criterion, but complex to
• calculate

34
Selection

• Based on fitness, choose the set of individuals
  (the intermediate population) to
• survive untouched, or
• be mutated, or
• in pairs, be crossed over and possibly mutated
• forming the next population
• One individual may be appear several times in the
  intermediate population (or the next population)

35
Types of Selection

• Using relative fitness (examples)
• roulette wheel -- classical Holland -- chunk of
  wheel relative fitness
• stochastic uniform sampling -- better sampling --
  integer parts GUARANTEED
• Not requiring relative fitness
• tournament selection
• rank-based selection (proportional or cutoff)
• elitist (mu, lambda) or (mulambda) from ES

36
Scaling of Relative Fitnesses

• Trouble as evolution progresses, relative
  fitness differences get smaller (as population
  gets more similar to each other). Often helpful
  to SCALE relative fitnesses to keep about same
  ratio of best guy/average guy, for example.
• Even better use tournament or rank-based or
  elitist selection

37
Explaining Why a GA Works Intro to GA Theory

• Some classical results
• Schema theorem how search effort is allocated
• Implicit parallelism each evaluation provides
  information on many possible candidate solutions
• k-Armed Bandit problem

38
What is a GA DOING?-- Schemata and Hyperstuff

• Schema -- adds to alphabet, means dont
  care any value
• One schema, two schemata (forgive occasional
  misuse in Whitley)
• Definition ORDER of schema H -- o(H) of
  non-s
• Def. Defining Length of a schema, D(H)
  distance between first and last non- in a
  schema for example D (1010) 5
  ( number of positions where 1-pt crossover can
  disrupt it).
• (NOTE diff. xover ? diff. relationship to
  defining length)
• Strings or chromosomes or individuals or
  solutions are order L schemata, where L is
  length of chromosome (in bits or loci).
  Chromosomes are INSTANCES (or members) of
  lower-order schemata

39
Cube and Hypercube
Vertices are order ? schemata Edges are order ?
schemata Planes are order ? schemata Cubes (a
type of hyperplane) are order ? schemata 8
different order-1 schemata (cubes) 0, 1,
0, 1, 0, 1, 0, 1
40
Hypercubes, Hyperplanes, etc.

• (See pictures in Whitley tutorial or blackboard)
• Vertices correspond to order L schemata (strings)
• Edges are order L-1 schemata, like 10 or 101
• Faces are order L-2 schemata
• Etc., for hyperplanes of various orders
• A string is an instance of 2L-1 schemata or a
  member of that many hyperplane partitions (-1
  because all s, the whole space, is not
  counted as a schema, per Holland)
• List them, for L3

41
GA Sampling of Hyperplanes

• So, in general, string of length L is an instance
  of 2L-1 schemata
• But how many schemata are there in the whole
  search space?
• (how many choices each locus?)
• Since one string instances 2L-1 schemata, how
  much does a population tell us about schemata of
  various orders?
• Implicit parallelism one strings fitness tells
  us something about relative fitnesses of more
  than one schema.

42
Fitness and Schema/ Hyperplane Sampling
Whitleys illustration of various partitions of
fitness hyperspace Plot fitness versus one
variable discretized as a K 4-bit binary
number then get ? First graph shades 0 Second
superimposes 1, so crosshatches are ? Third
superimposes 010
43
How Do Schemata Propagate? Proportional Selection
Favors Better Schemata

• Select the INTERMEDIATE population, the parents
  of the next generation, via fitness-proportional
  selection
• Let M(H,t) be number of instances (samples) of
  schema H in population at time t. Then
  fitness-proportional selection yields an
  expectation of
• In an example, actual number of instances of
  schemata (next page) in intermediate generation
  tracked expected number pretty well, in spite of
  small pop size

44
Results of example run (Whitley) showing that
observed numbers of instances of schemata track
expected numbers pretty well
45
Crossover Effect on Schemata

• One-point Crossover Examples (blackboard)
• 11 and 11
• Two-point Crossover Examples (blackboard)
• (rings)
• Closer together loci are, less likely to be
  disrupted by crossover, right? A compact
  representation is one that tends to keep alleles
  together under a given form of crossover
  (minimizes probability of disruption).

46
Linkage and Defining Length

• Linkage -- coadapted alleles (generalization of
  a compact representation with respect to
  schemata)
• Example, convincing you that probability of
  disruption of schema H of length D(H) is
  D(H)/(L-1)

47
The Fundamental Theorem of Genetic Algorithms --
The Schema Theorem

• Holland published in 1975, had taught it much
  earlier (by 1968, for example, when I started
  Ph.D. at UM)
• It provides lower bound on change in sampling
  rate of a single schema from generation t to t1.
  Well derive it in several steps, starting from
  the change caused by selection alone

48
Schema Theorem Derivation (cont.)

• Now we want to add effect of crossover
• A fraction pc of pop undergoes crossover, so
• Will make a conservative assumption that
  crossover within the defining length of H is
  always disruptive to H, and will ignore gains
  (were after a LOWER bound -- wont be as tight,
  but simpler). Then

49
Schema Theorem Derivation (cont.)

• Whitley considers one non-disruption case that
  Holland didnt, originally
• If cross H with an instance of itself, anywhere,
  get no disruption. Chance of doing that, drawing
  second parent at random, is P(H,t)
  M(H,t)/popsize so prob. of disruption by x-over
  is
• Then can simplify the inequality, dividing by
  popsize and rearranging re pc
• This version ignores mutation and assumes second
  parent is chosen at random. But its usable,
  already!

50
Schema Theorem Derivation (cont.)

• Now, lets recognize that well choose the second
  parent for crossover based on fitness, too
• Now, lets add mutations effects. What is the
  probability that a mutation affects schema H?
• (Assuming mutation always flips bit or changes
  allele)
• Each fixed bit of schema (o(H) of them) changes
  with probability pm, so they ALL stay UNCHANGED
  with probability

51
Schema Theorem Derivation (cont.)

• Now we have a more comprehensive schema theorem
• (This is where Whitley stops. We can use this
  but)
• Holland earlier generated a simpler, but less
  accurate bound, first approximating the mutation
  loss factor as (1-o(H)pm), assuming pmltlt1.

52
Schema Theorem Derivation (cont.)

• That yields
• But, since pmltlt1, we can ignore small
  cross-product terms and get
• That is what many people recognize as the
  classical form of the schema theorem.
• What does it tell us?

53
Using the Schema Theorem

• Even a simple form helps balance initial
  selection pressure, crossover mutation rates,
  etc.
• Say relative fitness of H is 1.2, pc .5, pm
  .05 and L 20 What happens to H, if H is long?
  Short? High order? Low order?
• Pitfalls slow progress, random search,
  premature convergence, etc.
• Problem with Schema Theorem important at
  beginning of search, but less useful later...

54
Building Block Hypothesis

• Define a Building block as a short, low-order,
  high-fitness schema
• BB Hypothesis Short, low-order, and highly fit
  schemata are sampled, recombined, and resampled
  to form strings of potentially higher fitness we
  construct better and better strings from the best
  partial solutions of the past samplings.
• -- David Goldberg, 1989
• (GAs can be good at assembling BBs, but GAs
  are also useful for many problems for which BBs
  are not available)

55
Lessons (Not Always Followed)

• For newly discovered building blocks to be
  nurtured (made available for combination with
  others), but not allowed to take over population
  (why?)
• Mutation rate should be
  (but contrast with SA, ES, (1l),
  )
• Crossover rate should be
• Selection should be able to
• Population size should be (oops what can we say
  about this? so far)

56
A Traditional Way to Do GA Search

• Population large
• Mutation rate (per locus) 1/L
• Crossover rate moderate (lt0.3)
• Selection scaled (or rank/tournament, etc.) such
  that Schema Theorem allows new BBs to grow in
  number, but not lead to premature convergence

57
Schema Theorem and Representation/Crossover Types

• If we use a different type of representation or
  different crossover operator
• Must formulate a different schema theorem, using
  same ideas about disruption of schemata.
• See Whitley (Fig. 4) for paths through search
  space under crossover

58
Uniform Crossover Linkage

• 2-pt crossover is superior to 1-point
• Uniform crossover chooses allele for each locus
  at random from either parent
• Uniform crossover is thus more disruptive than
  1-pt or 2-pt crossover
• BUT uniform is unbiased relative to linkage
• If all you need is small populations and a rapid
  scramble to find good solutions, uniform xover
  sometimes works better but is this what you
  need a GA for? Hmmmm
• Otherwise, try to lay out chromosome for good
  linkage, and use 2-pt crossover (or Bookers 1987
  reduced surrogate crossover, (described below))

59
Inversion An Idea to Try to Improve Linkage

• Tries to re-order loci on chromosome BUT NOT
  changing meaning of loci in the process
• Means must treat each locus as (index, value)
  pair. Can then reshuffle pairs at random, let
  crossover work with them in order APPEAR on
  chromosome, but fitness function keep association
  of values with indices of fields, unchanging.

60
Classical Inversion Operator

• Example reverses field pairs i through k on
  chromosome
• (a,va), (b,vb), (c,vc), (d,vd), (e,ve), (f, vf),
  (g,vg)
• After inversion of positions 2-4, yields
• (a,va), (d,vd), (c,vc), (b,vb), (e,ve), (f, vf),
  (g,vg)
• Now fields a,d are more closely linked, 1-pt or
  2-pt crossover less likely to separate them
• In practice, seldom used must run problem for
  an enormous time to have such a second-level
  effect be useful. Need to do on population level
  or tag each inversion pattern (and force mates to
  have matching tags) or do repairs to crossovers
  to keep chromosomes legal i.e., possess one
  pair of each type.

61
Inversion NOT a Reordering Operator

• In contrast, if trying to solve for the best
  permutation of 0,N, use other reordering
  crossovers well discuss later. Thats NOT
  inversion!

62
Crossover Between Similar Individuals

• As search progresses, more individuals tend to
  resemble each other
• When two similar individuals are crossed, chances
  of yielding children different from parents are
  lower for 1,2-pt than uniform
• Can counter this with reduced surrogate
  crossover (1-pt, 2-pt)

63
Reduced Surrogates

• Given 0001111011010011 and
• 0001001010010010, drop matching
• Positions, getting
• ----11---1-----1 and
• ----00---0-----0, reduced surrogates
• If pick crossover pts IGNORING DASHES, 1-pt, 2-pt
  still search similarly to uniform.

64
The Case for Binary Alphabets

• Deals with efficiency of sampling schemata
• Minimal alphabet ? maximum hyperplanes directly
  available in encoding, for schema processing and
  higher rate of sampling low-order schemata than
  with larger alphabet
• (See p. 20, Whitley, for tables)
• Half of a random init. pop. samples each order 1
  schema, and ¼ samples each order-2 schema, etc.
• If use alpha_size 10, many schemata of order 2
  will not be sampled in an initial population of
  50. (Of course, each order-1 schema sampled gave
  us info about a 3-bit allele

65
Case Against

• Antonisse raises counter-arguments on a
  theoretical basis, and the question of
  effectiveness is really open.
• But, often dont want to treat chromosome as bit
  string, but encode ints, allow crossover only
  between int fields, not at bit boundaries, use
  problem-specific representations.
• Losses in schema search efficiency may be
  outweighed by gains in naturalness of mapping,
  keeping fields legal, etc.
• So we will most often use non-binary strings
• (GALOPPS lets you go either way)

66
The N3 Argument (Implicit or Intrinsic
Parallelism)

• Assertion A GA with pop size N can usefully
  process on the order of N3 hyperplanes (schemata)
  in a generation.
• (WOW! If N100, N3 1 million)
• Derivation -- Assume
• Random population of size N.
• Need f instances of a schema to claim we are
  processing it in a statistically significant
  way in one generation.

67
The N3 Argument (cont.)

• Example to have 8 samples (on average) of 2nd
  order schemata in a pop., (there are 4 distinct
  (CONFLICTING) schemata in each 2-position pair
  for example, 00, 01, 10, 11),
  wed need 4 bit patterns x 8 instances 32
  popsize.
• In general, the highest ORDER of schema, ,
  that is processed is log (N/f) in our case,
  log(32/8) log(4) 2. (log means log2)

68
The N3 Argument (cont.)

• But the number of distinct schemata of order
• is , the number of ways to pick
  different positions and assign all possible
  binary values to each subset of the positions.
  
• So we are trying to argue that ,
  
• which implies that ,
  since
• log(N/f).

69
The N3 Argument (cont.)

• Rather than proving anything general, Fitzpatrick
  Grefenstette (88) argued as follows
• Assume
• Pick f8, which implies
• By inspection (plug in Ns, get s, etc.), the
  number of schemata processed is greater than N3.
  So, as long as our population size is REASONABLE
  (64 to a million) and L is large enough (problem
  hard enough), the argument holds.
• But this deals with the initial population, and
  it does not necessarily hold for the latter
  stages of evolution. Still, it may help to
  explain why GAs can work so well

70
Exponentially Increasing Sampling and the K-Armed
Bandit Problem

• Schema Theorem says M(H,t1) gt k M(H,t)
• (if we neglect certain changes)
• That is, Hs instances in population grow
  exponentially, as long as small relative to pop
  size and kgt1 (H is a building block).
• Is this a good way to allocate trials to
  schemata? Argument that SHOULD devote
  exponentially increasing fraction of trials to
  schemata that have performed better in samples so
  far

71
Two-Armed Bandit Problem(from Goldberg, 89)

• 1-armed bandit slot machine
• 2-armed bandit slot machine with 2 handles, NOT
  necessarily yielding same payoff odds (2
  different slot machines)
• If can make a total of N pulls, how should we
  proceed, so as to maximize expected final total
  payoff Ideas???

72
Two-Armed Bandit, cont.

• Assume LEFT pays with (unknown to us) expected
  value m1 and variance s12, and RIGHT pays m2,
  with variance s22.
• The DILEMMA Must EXPLORE while EXPLOITING.
  Clearly a tradeoff must be made. Given that one
  arm seems to be paying off better than the other
  SO FAR, how many trials should be given to the
  BETTER (so far) arm, and how many to the POORER
  (so far) arm?

73
Two-Armed Bandit, cont.

•  Classical approach SEPARATE EXPLORATION from
  EXPLOITATION If will do N trials, start by
  allocating n trials to each arm (2nltN) to decide
  WHICH arm appears to be better, and then allocate
  ALL remaining (N-2n) trials to it.
• DeJong calculated the expected loss (compared to
  the OPTIMUM) of using this strategy
• L(N,n) m1 - m2 . (N-n) q(n)
  n(1-q(n)),where q(n) is the probability that the
  WORST arm is the OBSERVED BEST arm after n trials
  on each machine.

74
Two-Armed Bandit, cont.

• This q(n) is well approximated by the tail of the
  normal distribution
• , where
• (x is signal difference to noise ratio times
  sqrt(n).)
• (Lets call signal difference to noise ratio c.)

q(n)
x
75
Two-Armed Bandit, cont.

• The LARGER x becomes, the LESS probable q(n)
  becomes (i.e., smaller chance of error). You can
  see that q(n) (chance of error) DECLINES as n is
  picked larger, or as the differences in expected
  values INCREASES or as the sum of the variances
  DECREASES.
• The equation shows two sources of expected loss
• L(N,n) m1 - m2 . (N-n) q(n) n(1-q(n)),
• Due to wrong arm later
  wrong during exploration

76
Two-Armed Bandit, cont.

• For any N, solve for the optimal experiment size
  n by setting the derivative of the loss equation
  to 0. Graph below (after Fig. 2.2 in Goldberg,
  89) shows the optimal n as a function of total
  number of trials, N, and c, the ratio of signal
  difference to noise.

From graph, see that total number of experiments
N grows at a greater-than-exponential function of
the ideal number of trials n in the exploration
period -- that means, according to classical
decision theory, that we should be allocating
trials to the BETTER (higher measured fitness
during the exploration period) of the two arms,
at a GREATER THAN EXPONENTIAL RATE.
77
Two-Armed Bandit, K-Armed Bandit

• Now, let our arms represent competing schemata.
  Then the future sampling of the better one (to
  date) should increase at a larger-than-exponential
  rate. A GA, using selection, crossover, and
  mutation, does that (when set properly, according
  to the schema theorem). If there are K competing
  schemata over a set of positions, then its a
  K-armed bandit.
• But at any time, MANY different schemata are
  being processed, with each competing set
  representing a K-armed bandit scenario. So maybe
  the GAs way of allocating trials to schemata is
  pretty good!

78
Early Theory for GAs

• Vose and Liepins (91) produced most well-known
  GA theory model
• The main elements
• vector of size 2L containing proportion of
  population with genotype i at time t (before
  selection), P(Si,t), whole vector denoted pt,
• matrix rij(k) of probabilities that crossing
  strings i and j will produce string k.
• Then

79
Vose Liepins (cont.)

• r is used to construct M, the mixing matrix
  that tells, for each possible string, the
  probability that it is created from each pair of
  parent strings. Mutation can also be included to
  generate a further huge matrix that, in theory,
  could be used, with an initial population, to
  calculate each successive step in evolution.

80
Vose Liepins (cont.)

• The problem is that not many theoretical results
  with practical implications can be obtained,
  because for interesting problems, the matrices
  are too huge to be usable, and the effects of
  selection are difficult to estimate. More recent
  work in a statistical mechanics approach to GA
  theory seems to me to hold far more interest.

81
What are Common Problems when Using GAs in
Practice?

• Hitchhiking BB1.BB2.junk.BB3.BB4 junk adjacent
  to building blocks tends to get fixed can be
  a problem
• Deception a 3-bit deceptive function
• Epistasis nonlinear effects, more difficult to
  capture if spread out on chromosome

82
In PRACTICE GAs Do a JOB

• DOESNT mean necessarily finding global optimum
• DOES mean trying to find better approximate
  answers than other methods do, within the time
  available!
• People use any dirty tricks that work
• Hybridize with local search operations
• Use multiple populations/multiple restarts, etc.
• Use problem-specific representations and
  operators
• The GOALS
• Minimize of function evaluations needed
• Balance exploration/exploitation so get best
  answer can during time available (AVOIDING
  premature convergence)

83
Different Forms of GA

• Generational vs. Steady-State
• Generation gap 1.0 means replace ALL by newly
  generated children at lower extreme, generate
  1 (or 2) offspring per generation (called
  steady-state)
• (GALOPPS allows either, by setting crossover
  rates)

84
Different Forms of GA

• Replacement Policy
• Offspring replace parents
• K offspring replace K worst ones
• Offspring replace random individuals in
  intermediate population
• Offspring are crowded in
• (GALOPPS allows 1,3,4 easily, 2 takes mods)

85
Crowding

• Crowding (DeJong) helps form niches and avoid
  premature takeover by fit individuals
• For each child
• Pick K candidates for replacement, at random,
  from intermediate population
• Calculate pseudo-Hamming distance from child to
  each
• Replace individual most similar to child
• Effect?

86
Elitism

• Artificially protects fittest K members of
  population against replacement in next generation
• Often useful, but beware if using multiple
  subpopulations
• K often 1 may be larger, even large
• (ES often keeps k best of offspring, or of
  offspring and parents, throws away the rest)

87
Example GA Packages GENITOR (Whitley)

• Steady-state GA
• Child replaces worst-fit individual
• Fitness is assigned according to rank (so no
  scaling is needed)
• (elitism is automatic)
• (Can do in GALOPPS except worst replacement
  user must rewrite that part)

88
Example GA Packages CHC (Eshelman)

• Elitism -- (ml) from ES generate l offspring
  from m parents, keep best m of the ml parents
  and children.
• Uses incest prevention (reduction) pick mates
  on basis of their Hamming dissimilarity
• HUX form of uniform crossover, highly
  disruptive
• Rejuvenate with cataclysmic mutation when
  population starts converging, which is often
  (small populations used)
• GALOPPS allows last three, not first one
• I dont favor except for relatively easy problem
  spaces

89
Hybridizing GAs a Good Idea!

• IDEA combine a GA with local or
  problem-specific search algorithms
• HOW typically, for some or all individuals,
  start from GA solution, take one or more steps
  according to another algorithm, use resulting
  fitness as fitness of chromosome.
• If also change genotype, Lamarckian if dont,
  Baldwinian (preserves schema processing)
• Helpful in many constrained optimization problems
  to repair infeasible solutions to nearby
  feasible ones

90
Other Representations/OperatorsPermutation/Optim
al Ordering

• Chromosome has EXACTLY ONE copy of each int in
  0,N-1
• Must find optimal ordering of those ints
• 1-pt, 2-pt, uniform crossover ALL useless
• Mutations swap 2 loci, scramble K adjacent
  loci, shuffle K arbitrary loci, etc.
• (See blackboard for example)

91
Crossover Operators for Permutation Problems

• What properties do we want
• 1) Want each child to combine building blocks
  from both parents in a way that preserves
  high-order schemata in as meaningful a way as
  possible, and
• 2) Want all solutions generated to be feasible
  solutions.

92
Example Operators for Permutation-Based
Representations, Using TSP Example PMX --
Partially Matched Crossover

• 2 sites picked, intervening section specifies
  cities to interchange between parents
• A 9 8 4 5 6 7 1 3 2 10
• B 8 7 1 2 3 10 9 5 4 6
• A 9 8 4 2 3 10 1 6 5 7
• B 8 10 1 5 6 7 9 2 4 3
• (i.e., swap 5 with 2, 6 with 3, and 7 with 10 in
  both children.)
• Thus, some ordering information from each parent
  is preserved, and no infeasible solutions are
  generated.

93
Example Operators forPermutation-Based
Representations Order Crossover

• A 9 8 4 5 6 7 1 3 2 10 (segment A
  and B)
• B 8 7 1 2 3 10 9 5 4 6
• gt B 8 H 1 2 3 10 9 H 4 H (repl. 5 6 7
  with Hs)
• gt B 2 3 10 H H H 9 4 8 1 (promote
  segment from B, gather Hs, append rest, with
  wrap-around)
• gt B 2 3 10 5 6 7 9 4 8 1
• Similarly, A 5 6 7 2 3 10 1 9 8 4
• Order crossover preserves more information about
  RELATIVE ORDER than does PMX, but less about
  ABSOLUTE POSITION of each city (for TSP
  example).

94
Example Operators forPermutation-Based
Representations Cycle Crossover

• Cycle crossover forces the city in each position
  to come from that same position on one of the two
  parents
• C 9 8 2 1 7 4 5 10 6 3
• D 1 2 3 4 5 6 7 8 9 10
• 9 - - - - - - - - -
• gt 9 - - 1 - - - - - -
• gt 9 - - 1 - 4 - - 6 - , which completes 1st
  cycle then (depending on whose cycle crossover
  you choose), (i) start from first unassigned
  position in D and perform another cycle, or (ii)
  just fill in the rest of the numbers from
  chromosome D
• (i) yields gt 9 2 - 1 - 4 - 8 6 10
• gt 9 2 3 1 - 4 - 8 6 10
• gt C 9 2 3 1 7 4 5 8 6 10
  D is done similarly.
• (ii) yields gt C 9 2 3 1 5 4 7 8 6 10. D
  is done similarly.

95
Example Operators forPermutation-Based
Representations Uniform Order-Based Crossover

• ( lt Lawrence Davis, Handbook of Genetic
  Algorithms)
• Analogous to uniform crossover for ordinary
  list-based chromosomes. Uniform crossover
  effectively acts as if many one- or two-point
  crossovers were performed at once on a pair of
  chromosomes, combining parents genes on a
  locus-by-locus basis, so is quite disruptive of
  longer schemata. (I dont like it much, as it
  jumbles information and is too disruptive for
  effectiveness with many problems, I believe. But
  it works quite well for some others.)
• A 1 2 3 4 5 6 7 8
• B 8 6 4 2 7 5 3 1
• Binary Template 0 1 1 0 1 1 0 0
  (random)
• gt - 2 3 - 5 6 - -
• (then, reordering rest of As nodes to the order
  THEY appear in B)
• gt A 8 2 3 4 5 6 7 1
• (and similarly for B, gt 8 4 5 2 6 7 3 1

96
Parallel GAs Independent of Hardware

• Three primary models coarse-grain (island),
  fine-grain (cellular), and micro-grain (trivial)
• Trivial (not really a parallel GA just a
  parallel implementation of a single-population
  GA) pass out individuals to separate processors
  for evaluation (or run lots of local tournaments,
  no master) still acts like one large population

97
Coarse-Grain (Island) Parallel GA

• N independent subpopulations, acting as if
  running in parallel (timeshared or actually on
  multiple processors)
• Occasionally, migrants go from one to another, in
  pre-specified patterns
• Strong capability for avoiding premature
  convergence while exploiting good individuals, if
  migration rates/patterns well chosen

98
GALOPPS An Island Parallel GA

• Can run 1-99 subpopulations
• Can run all in one process
• Can run any number in separate processes on one
  uni- or multi-processor
• Can run any number of subpopulations on each of K
  processors need only share a common DISK
  directory

99
Migrant Selection Policy

• Who should migrate?
• Best guy?
• One random guy?
• Best and some random guys?
• Guy very different from best of receiving subpop?
  (incest reduction)
• If migrate in large of population each
  generation, acts like one big population, but
  with extra replacements could actually SPEED
  premature convergence

100
Migrant Replacement Policy

• Who should a migrant replace?
• Random individual?
• Worst individual?
• Most similar individual (Hamming sense)
• Similar individual via crowding?

101
How Many Subpopulations?(Crude Rule of Thumb)

• How many total evaluations can you afford?
• Total population size and number of generations
  and generation gap determine run time
• What should minimum subpopulation size be?
• Smaller than 40-50 USUALLY spells trouble rapid
  convergence of subpop 100-200 better for some
  problems
• Divide to get how many subpopulations you can
  afford

102
Fine-Grain Parallel GAs

• Individuals distributed on cells in a
  tessellation, one or few per cell (often,
  toroidal checkerboard)
• Mating typically among near neighbors, in some
  defined neighborhood
• Offspring typically placed near parents
• Can help to maintain spatial niches, thereby
  delaying premature convergence
• Interesting to view as a cellular automaton

103
Refined Island Models Heterogeneous/
Hierarchical GAs

• For many problems, useful to use different
  representations/levels of refinement/types of
  models, allow them to exchange nuggets
• GALOPPS was first package to support this
• Injection Island architecture arose from this,
  now used in HEEDS, etc.
• Hierarchical Fair Competition is newest
  development (Jianjun Hu), breaking populations by
  fitness bands

104
Multi-Level GAs

• Pioneering Work DAGA2, MSU (based on GALOPPS)
• Island GA populations are on lower level, their
  parameters/operators/ neighborhoods on chromosome
  of a single higher-level population that controls
  evolution of subpopulations
• Excellent performance reproducible trajectories
  through operator space, for example

105
Examples of Population-to-Population Differences
in a Heterogeneous GA

• Different GA parameters (pop size, crossover
  type/rate, mutation type/rate, etc.)
• 2-level or without a master pop
• Examples of Representation Differences
• Hierarchy one-way migration from least refined
  representation to most refined
• Different models in different subpopulations
• Different objectives/constraints in different
  subpops (sometimes used in Evolutionary
  Multiobjective Optimization (EMOO)) (someone
  pick an EMOO paper?)

106
Additional GA Topics to Come

• EMOO Evolutionary Multi-Objective Optimization
• Differential Evolution GA with a twist
• PCX Parent-Centered Crossover
• CMA-ES? (maybe)

107
Evolutionary Multi-Objective Optimisation

• EMOO Evolutionary Multi-Objective Optimization
  (sometimes, MOGA, MOEA)
• Many well-known methods VEGA, NPGA, NSGA, SPEA,
  NSGA-II
• Excellent books by Deb and by Coello-Coello

108
Multi-Objective Optimization Problem (Constrained)
If g(x) or h(x) violated, solution is INFEASIBLE.
109
Non-Dominated Solutions
(text from http//ieeexplore.ieee.org/iel5/20/2150
0/00996290.pdf)
110
Non-Dominated Setsand Pareto Sets

• So a solution thats same or worse on all
  objectives than some other solution is dominated
  by that solution else its non-dominated by that
  solution.
• A set consisting only of non-dominated points is
  a Pareto set or non-dominated set (i.e., no point
  in the set dominates any other) sometimes also
  used in sense of points not dominated by any
  other points already visited in the search space,
  or as the non-dominated subset of a larger set of
  points.
• THE set of ALL points that are not dominated by
  any other feasible solutions in the space is
  called THE Pareto Front (it is unique).

111
How Should a Constrained MOGA, Do its Work?

• What would you do? (Try it on blackboard)
• want lots of solutions
• want to approximate Pareto front
• want them well distributed along front
• want them to satisfy constraints

112
How Does NSGA-II (Deb) Do It?

• Non-classical GA operators crossover and
  mutation, but thats not the question
• How determine fitness?
• Non-dominated sorting
• Double-sized intermediate population
• Constraints

113
Differential Evolution A Funny-Looking GA

• Well later look at the DE tutorial paper by Craft

114
How Do GAs Go Bad?

• Premature convergence
• Unable to overcome deception
• Need more evaluations than time permits
• Bad match of representation/mutation/crossover,
  making operators destructive
• Biased or incomplete representation
• Problem too hard
• (Problem too easy, makes GA look bad)

115
So, in Conclusion

• GAs can be easy to use, but not necessarily easy
  to use WELL
• Dont use them if something else will work it
  will probably be faster
• GAs cant solve every problem, either
• GAs are only one of several strongly related
  branches of evolutionary computation and they
  all commonly get hybridized