# Genetic AIgorithms

Chapter 3
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
GA Quick Overview
eveloped: USA in the 1970's
Early names: J. Holland, K. eJong, . Goldberg
Typically applied to:
÷ discrete optimization
Attributed features:
÷ not too fast
÷ good heuristic for combinatorial problems
Special Features:
÷ Traditionally emphasizes combining information from good
parents (crossover)
÷ many variants, e.g., reproduction models, operators
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Genetic aIgorithms
Holland's original GA is now known as the
simple genetic algorithm (SGA)
Other GAs use different:
÷ Representations
÷ Mutations
÷ Crossovers
÷ Selection mechanisms
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
SGA technicaI summary tabIeau
Representation Binary strings
Recombination N-point or uniform
Mutation Bitwise bit-flipping with fixed
probability
Parent selection Fitness-Proportionate
Survivor selection All children replace parents
Speciality Emphasis on crossover
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Genotype space =
{0,1}
L
Phenotype space
Encoding
(representation)
ecoding
(inverse representation)
011101001
010001001
10010010
10010001
Representation
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
SGA reproduction cycIe
1. Select parents for the mating pool
(size of mating pool = population size)
2. Shuffle the mating pool
3. For each consecutive pair apply crossover with
probability p
c
, otherwise copy parents
4. For each offspring apply mutation (bit-flip with
probability p
m
independently for each bit)
5. Replace the whole population with the resulting
offspring
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
SGA operators: 1-point crossover
Choose a random point on the two parents
Split parents at this crossover point
Create children by exchanging tails
P
c
typically in range (0.6, 0.9)
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
SGA operators: mutation
Alter each gene independently with a probability p
m
p
m
is called the mutation rate
÷ Typically between 1/pop_size and 1/ chromosome_length
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Main idea: better individuals get higher chance
÷ Chances proportional to fitness
÷ Ìmplementation: roulette wheel technique
Assign to each individual a part of the
roulette wheel
Spin the wheel n times to select n
individuals
SGA operators: SeIection
fitness(A) = 3
fitness(B) = 1
fitness(C) = 2
A
C
1/6 = 17%
3/6 = 50%
B
2/6 = 33%
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
An exampIe after GoIdberg '89 (1)
Simple problem: max x
2
over {0,1,.,31}
GA approach:
÷ Representation: binary code, e.g. 01101 ÷13
÷ Population size: 4
÷ 1-point xover, bitwise mutation
÷ Roulette wheel selection
÷ Random initialisation
e show one generational cycle done by hand
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
x
2
exampIe: seIection
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
X
2
exampIe: crossover
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
X
2
exampIe: mutation
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
The simpIe GA
Has been subject of many (early) studies
÷ still often used as benchmark for novel GAs
Shows many shortcomings, e.g.
÷ Representation is too restrictive
÷ Mutation & crossovers only applicable for bit-string &
integer representations
÷ Selection mechanism sensitive for converging
populations with close fitness values
÷ Generational population model (step 5 in SGA repr.
cycle) can be improved with explicit survivor selection
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
AIternative Crossover Operators
Performance with 1 Point Crossover depends on the
order that variables occur in the representation
÷ more likely to keep together genes that are near
each other
÷ Can never keep together genes from opposite ends
of string
÷ This is known as Positional Bias
÷ Can be exploited if we know about the structure of
our problem, but this is not usually the case
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
n-point crossover
Choose n random crossover points
Split along those points
Glue parts, alternating between parents
Generalisation of 1 point (still some positional bias)
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Uniform crossover
Assign 'heads' to one parent, 'tails' to the other
Flip a coin for each gene of the first child
Make an inverse copy of the gene for the second child
Ìnheritance is independent of position
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Crossover OR mutation?
ecade long debate: which one is better / necessary /
main-background
Answer (at least, rather wide agreement):
÷ it depends on the problem, but
÷ in general, it is good to have both
÷ both have another role
÷ mutation-only-EA is possible, xover-only-EA would not work
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Exploration: iscovering promising areas in the search
space, i.e. gaining information on the problem
Exploitation: Optimising within a promising area, i.e. using
information
There is co-operation AN competition between them
somewhere "in between¨ two (parent) areas
Mutation is exploitative, it creates random small
diversions, thereby staying near (in the area of ) the parent
Crossover OR mutation? (cont'd)
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Only crossover can combine information from two
parents
Only mutation can introduce new information (alleles)
Crossover does not change the allele frequencies of
the population (thought experiment: 50% 0's on first
bit in the population, ?% after performing n
crossovers)
To hit the optimum you often need a 'lucky' mutation
Crossover OR mutation? (cont'd)
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Other representations
Gray coding of integers (still binary chromosomes)
÷ Gray coding is a mapping that means that small changes in
the genotype cause small changes in the phenotype (unlike
binary coding). "Smoother¨ genotype-phenotype mapping
makes life easier for the GA
Nowadays it is generally accepted that it is better to
encode numerical variables directly as
Ìntegers
Floating point variables
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Integer representations
Some problems naturally have integer variables, e.g.
image processing parameters
Others take categorical values from a fixed set e.g.
{blue, green, yellow, pink}
N-point / uniform crossover operators work
Extend bit-flipping mutation to make
÷ "creep¨ i.e. more likely to move to similar value
÷ Random choice (esp. categorical variables)
÷ For ordinal problems, it is hard to know correct range for
creep, so often use two mutation operators in tandem
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
ReaI vaIued probIems
Many problems occur as real valued problems, e.g.
continuous parameter optimisation f : T
n
T
Ìllustration: Ackley's function (often used in EC)
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Mapping reaI vaIues on bit strings
z x,y] _ Trepresented by {a
1
,.,a
L
} {0,1}
L
· x,y] ÷{0,1}
L
must be invertible (one phenotype per
genotype)
· I: {0,1}
L
÷x,y] defines the representation
Only 2
L
values out of infinite are represented
L determines possible maximum precision of solution
High precision long chromosomes (slow evolution)
| , | ) 2 (
1 2
) ,..., (
1
0
1
,

, ,

+ = I
¯

=

A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
FIoating point mutations 1
General scheme of floating point mutations
Uniform mutation:
Analogous to bit-flipping (binary) or random resetting
(integers)

@ @
=
@
÷ = ..., , ..., ,
1 1
)

& , ,
@
)

& , Irom (uniIorm) randomly drawn @
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
FIoating point mutations 2
Non-uniform mutations:
÷ Many methods proposed,such as time-varying
range of change etc.
÷ Most schemes are probabilistic but usually only
make a small change to value
÷ Most common method is to add random deviate to
each variable separately, taken from N(0, 9)
Gaussian distribution and then curtail to range
÷ Standard deviation 9 controls amount of change
(2/3 of deviations will lie in range (- 9 to + 9)
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Crossover operators for reaI vaIued GAs
iscrete:
÷ each allele value in offspring z comes from one of its
parents (x,y) with equal probability: z
i
= x
i
or y
i
÷ Could use n-point or uniform
Ìntermediate
÷ exploits idea of creating children "between¨ parents
(hence a.k.a. arithmetic recombination)
÷ z
i
= - x
i
+ (1 - -) y
i
where - : 0 A - A 1.
÷ The parameter - can be:
· constant: uniform arithmetical crossover
· variable (e.g. depend on the age of the population)
· picked at random every time
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
SingIe arithmetic crossover
· Parents: x
1
,.,x
n
and y
1
,.,y
n

· Pick a single gene () at random,
· child
1
is:
· reverse for other child. e.g. with - = 0.5
3
..., , ) 1 ( , ..., ,
1
+ - -
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
SimpIe arithmetic crossover
· Parents: x
1
,.,x
n
and y
1
,.,y
n

· Pick random gene () after this point mix values
· child
1
is:
· reverse for other child. e.g. with - = 0.5
3

+
+
+
+
) 1 (
n
y ..., ,
1
) 1 (
1
, ..., ,
1
- - - -
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
· Most commonly used
· Parents: x
1
,.,x
n
and y
1
,.,y
n

· child
1
is:
· reverse for other child. e.g. with - = 0.5
WhoIe arithmetic crossover
, , + ) 1 (
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Permutation Representations
Ordering/sequencing problems form a special type
Task is (or can be solved by) arranging some objects in
a certain order
÷ Example: sort algorithm: important thing is which elements
occur before others (order)
÷ Example: Travelling Salesman Problem (TSP) : important thing
is which elements occur next to each other (adjacency)
These problems are generally expressed as a
permutation:
÷ if there are n variables then the representation is as a list of n
integers, each of which occurs exactly once
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Permutation representation: TSP exampIe
Problem:
· Given n cities
· Find a complete tour with
minimal length
Encoding:
· Label the cities 1, 2, . , n
· One complete tour is one
permutation (e.g. for n =4
1,2,3,4], 3,4,2,1] are OK)
Search space is BÌG:
for 30 cities there are 30! 10
32
possible tours
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Mutation operators for permutations
solutions
÷ e.g. bit-wise mutation : let gene i have value j
÷ changing to some other value would mean that
occurred twice and j no longer occurred
Therefore must change at least two values
Mutation parameter now reflects the probability
that some operator is applied once to the
whole string, rather than individually in each
position
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Insert Mutation for permutations
Pick two allele values at random
Move the second to follow the first, shifting the
rest along to accommodate
Note that this preserves most of the order and
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Swap mutation for permutations
Pick two alleles at random and swap their
positions
Preserves most of adjacency information (4
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Inversion mutation for permutations
Pick two alleles at random and then invert the
substring between them.
breaks two links) but disruptive of order
information
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
ScrambIe mutation for permutations
Pick a subset of genes at random
Randomly rearrange the alleles in those
positions
(note subset does not have to be contiguous)
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
"Normal¨ crossover operators will often lead to
Many specialised operators have been devised
which focus on combining order or adjacency
information from the two parents
Crossover operators for permutations
1 2 3 4 5
5 4 3 2 1
1 2 3 2 1
5 4 3 4 5
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Order 1 crossover
Ìdea is to preserve relative order that elements occur
Ìnformal procedure:
1. Choose an arbitrary part from the first parent
2. Copy this part to the first child
3. Copy the numbers that are not in the first part, to
the first child:
starting right from cut point of the copied part,
using the order of the second parent
and wrapping around at the end
4. Analogous for the second child, with parent roles
reversed
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Order 1 crossover exampIe
Copy randomly selected set from first parent
Copy rest from second parent in order 1,9,3,8,2
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Ìnformal procedure for parents P1 and P2:
1. Choose random segment and copy it from P1
2. Starting from the first crossover point look for elements in that
segment of P2 that have not been copied
3. For each of these i look in the offspring to see what element j has
been copied in its place from P1
4. Place i into the position occupied j in P2, since we know that we will
not be putting j there (as is already in offspring)
5. Ìf the place occupied by j in P2 has already been filled in the
offspring , put i in the position occupied by in P2
6. Having dealt with the elements from the crossover segment, the rest
of the offspring can be filled from P2.
Second child is created analogously
PartiaIIy Mapped Crossover (PMX)
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
PMX exampIe
Step 1
Step 2
Step 3
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
CycIe crossover
Basic idea:
Each allele comes from one parent together with its position.
Ìnformal procedure:
1. Make a cycle of alleles from P1 in the following way.
(b) Look at the allele at the same position in P2.
(c) Go to the position with the same allele in P1.
(d) Add this allele to the cycle.
(e) Repeat step b through d until you arrive at the first allele of P1.
2. Put the alleles of the cycle in the first child on the positions
they have in the first parent.
3. Take next cycle from second parent
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
CycIe crossover exampIe
Step 1: identify cycles
Step 2: copy alternate cycles into offspring
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Edge Recombination
orks by constructing a table listing which
edges are present in the two parents, if an
edge is common to both, mark with a +
e.g. 1 2 3 4 5 6 7 8 9] and 9 3 7 8 2 6 5 1 4]
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Edge Recombination 2
Ìnformal procedure once edge table is constructed
1. Pick an initial element at random and put it in the offspring
2. Set the variable current element = entry
3. Remove all references to current element from the table
4. Examine list for current element:
÷ Ìf there is a common edge, pick that to be next element
÷ Otherwise pick the entry in the list which itself has the shortest list
÷ Ties are split at random
5. Ìn the case of reaching an empty list:
÷ Examine the other end of the offspring is for extension
÷ Otherwise a new element is chosen at random
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Edge Recombination exampIe
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
MuItiparent recombination
Recall that we are not constricted by the practicalities
of nature
Noting that mutation uses 1 parent, and "traditional¨
crossover 2, the extension to a2 is natural to examine
Been around since 1960s, still rare but studies indicate
useful
Three main types:
÷ Based on allele frequencies, e.g., p-sexual voting generalising
uniform crossover
÷ Based on segmentation and recombination of the parents, e.g.,
diagonal crossover generalising n-point crossover
÷ Based on numerical operations on real-valued alleles, e.g.,
center of mass crossover, generalising arithmetic
recombination operators
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
PopuIation ModeIs
SGA uses a Generational model:
÷ each individual survives for exactly one generation
÷ the entire set of parents is replaced by the offspring
At the other end of the scale are Steady-State
models:
÷ one offspring is generated per generation,
÷ one member of population replaced,
Generation Gap
÷ the proportion of the population replaced
÷ 1.0 for GGA, 1/pop_size for SSGA
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Fitness Based Competition
Selection can occur in two places:
÷ Selection from current generation to take part in
mating (parent selection)
÷ Selection from parents + offspring to go into next
generation (survivor selection)
Selection operators work on whole individual
÷ i.e. they are representation-independent
istinction between selection
÷ operators: define selection probabilities
÷ algorithms: define how probabilities are implemented
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
ImpIementation exampIe: SGA
Expected number of copies of an individual i
E( n
i
) = 3 · f(i)/ f
(3 = pop.size, f(i) = fitness of i, f avg. fitness in pop.)
Roulette wheel algorithm:
÷ Given a probability distribution, spin a 1-armed
wheel n times to make n selections
÷ No guarantees on actual value of n
i
Baker's SUS algorithm:
÷ n evenly spaced arms on wheel and spin once
÷ Guarantees floor(E( n
i
) ) A n
i
A ceil(E( n
i
) )
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Problems include
÷ One highly fit member can rapidly take over if rest of
population is much less fit: Premature Convergence
÷ At end of runs when fitnesses are similar, lose
selection pressure
÷ Highly susceptible to function transposition
Scaling can fix last two problems
÷ indowing: f'(i) = f(i) - .
t
where . is worst fitness in this (last n) generations
÷ Sigma Scaling: f'(i) = max( f(i) ÷ ( f - c · 9
f
), 0.0)
where c is a constant, usually 2.0
Fitness-Proportionate SeIection
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Function transposition for FPS
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Rank - Based SeIection
Attempt to remove problems of FPS by basing
selection probabilities on relative rather than
absolute fitness
Rank population according to fitness and then
base selection probabilities on rank where
fittest has rank 3 and worst rank 1
This imposes a sorting overhead on the
algorithm, but this is usually negligible
compared to the fitness evaluation time
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Linear Ranking
Parameterised by factor s: 1.0 < s A 2.0
÷ measures advantage of best individual
÷ in GGA this is the number of children allotted to it
Simple 3 member example
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
ExponentiaI Ranking
Linear Ranking is limited to selection pressure
Exponential Ranking can allocate more than 2
copies to fittest individual
Normalise constant factor c according to
population size
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Tournament SeIection
All methods above rely on global population
statistics
÷ Could be a bottleneck esp. on parallel machines
÷ Relies on presence of external fitness function
which might not exist: e.g. evolving game players
Ìnformal Procedure:
÷ Pick members at random then select the best of
these
÷ Repeat to select more individuals
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Tournament SeIection 2
Probability of selecting i will depend on:
÷ Rank of i
÷ Size of sample
higher increases selection pressure
÷ hether contestants are picked with replacement
Picking without replacement increases selection pressure
÷ hether fittest contestant always wins
(deterministic) or this happens with probability p
For = 2, time for fittest individual to take over
population is the same as linear ranking with s = 2 · p
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Survivor SeIection
Most of methods above used for parent
selection
Survivor selection can be divided into two
approaches:
÷ Age-Based Selection
e.g. SGA
Ìn SSGA can implement as "delete-random¨ (not
recommended) or as first-in-first-out (a.k.a. delete-oldest)
÷ Fitness-Based Selection
Using one of the methods above or
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Two SpeciaI Cases
Elitism
÷ idely used in both population models (GGA,
SSGA)
÷ Always keep at least one copy of the fittest solution
so far
GENÌTOR: a.k.a. "delete-worst¨
÷ From hitley's original Steady-State algorithm (he
also used linear ranking for parent selection)
÷ Rapid takeover : use with large populations or "no
duplicates¨ policy
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
ExampIe appIication of order based GAs: JSSP
Precedence constrained job shop scheduling problem
J is a set of jobs.
O is a set of operations
M is a set of machines
ble _ O L M defines which machines can perform which
operations
Pre _ O L O defines which operation should precede which
ur : _ O L M ÷ÌR defines the duration of o O on m M
The goal is now to find a schedule that is:
Complete: all jobs are scheduled
Correct: all conditions defined by ble and Pre are satisfied
Optimal: the total duration of the schedule is minimal
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
Precedence constrained job shop scheduIing GA
Representation: individuals are permutations of operations
Permutations are decoded to schedules by a decoding procedure
÷ take the first (next) operation from the individual
÷ look up its machine (here we assume there is only one)
÷ assign the earliest possible starting time on this machine, subject to
machine occupation
precedence relations holding for this operation in the schedule created so far
fitness of a permutation is the duration of the corresponding
schedule (to be minimized)
use any suitable mutation and crossover
use roulette wheel parent selection on inverse fitness
Generational GA model for survivor selection
use random initialisation
A.E. Eiben and J.E. Smith, Ìntroduction to Evolutionary Computing
Genetic Algorithms
JSSP exampIe: operator comparison

  -03,3/  \$29 3974/:.94394;4:943,7425:93 0309.47928 

":. ;07;0 
   

0;0450/&\$390 8 ,73,208 4,3/  043  4/-07 %5.,,550/94
/8.70904592,943 

997-:90/10,9:708
3499441,89 44/0:789.147.42-3,947,574-028 

\$50.,0,9:708
%7,/943,025,808.42-3331472,943174244/ 5,70398 .74884;07 2,3;,7,398 0  70574/:.94324/08 4507,9478

  -03,3/  \$29 3974/:.94394;4:943,7425:93 0309.47928 

0309.,47928 
 

4,3/ 8473,834343,890 82500309.,4792 \$ 9078:80/1107039
#05708039,9438 :9,9438 74884;078 \$00.94320.,3828

7425:93 0309.-0.3/  \$29 3974/:.94394..: #05708039..943 #0./703705.789738  543947:31472 980-9 1553910/ 574-.943 !.  -03.90 .47800.3..74884.8843.9 3.-9 93088 !74547943.79.70398 25..42-3.8:22.4:943.943 :9.943 \$50.07 .943 \$:7.05.47928 \$90.7039800.

3/  \$29 3974/:.  -03..4/3 705708039.943 !03495085.94394.4/3 3.0 3.0  <             0.4:943.943 03495085.943 .7425:93 0309.0780705708039.47928 #05708039..

.455.4380.3/  \$29 3974/:.-95. 490780.943.4118573.74884.70398  470.943990708:93 4118573 .70398147902.09040545:.:9.05.943 -9 159 574-.93544545:.55..7.93544 80412.  -03.94380  \$:110902..552:9...4:943.93544  470.47928 \$70574/:.7425:93 0309.-952 3/0503/0391470.079 574-..94394.0  \$00.-9  #05.95.

7425:93 0309.07     4480.30    .94394.90.339.4:943.95.3/  \$29 3974/:.70398 \$595.075439 70.9478 5439.3/4254394390945.  -03.47928 \$4507.74884./703-0.998.74884.7..8 !.37..70398.

.4:943..9437.574-.94394.94782:9.7425:93 0309.3/  \$29 3974/:.90 %5.47928 \$4507.-09003.  -03.0303/0503/0399..0/902:9.943   9070.-952 528.

3/ .545*80 .

.74248420*039 .

7425:93 0309..794190 74:099000  \$5390003920894800..3.36:0  883940.3/  \$29 3974/:.9478\$00./:.-099073/.80907.8 .47928 \$4507.  -03.3.94193088 2502039.08574547943./:.94374:09900090.4:943.3 /0.5./:.93 3/.3/.943  .94394..0 .

  .

    193088   193088   193088   .

 .

943-3.943.94394.3/  \$29 3974/:.943 #4:099000800..47928 30.7425:93 0309. 4.94380  54394.4:943.943 #.4/0 0      !45:.3/4239.250.07  .7.3/ .5574.8.. < #05708039.943  0844300307.07 -9802:9.  -03.19074/-07     \$250574-022..0/430-.

3/  \$29 3974/:.250800.94394.943 .7425:93 0309.4:943.  -03.47928  0.

47928  0.94394.4:943.74884.250.7425:93 0309.  -03.3/  \$29 3974/:.07 .

2502:9.94394.4:943.47928  0.943 .  -03.7425:93 0309.3/  \$29 3974/:.

.943.38479.07843.9412.74884.9438 \$00.08 \$482.8-0038:-0.:08 0307.943894470897.-0147-9 8973  39007705708039.47928 %08250   .9.94389.4:943.943 .943 .0147..47800..3/  \$29 3974/:.38280389.4238 0  #05708039.43.480193088..55.7425:93 0309.3-02574.  -03.714734.8-03.7 89:/08 8941903:80/.2.94324/0 89053\$7057  ...545:.073 545:.98:7.94320.0 :9.0 .3 0.94394.0/905.

90..8 .0700594090703081742455489003/8 418973 %88343.47928 9073.7425:93 0309..943 247009400594090703089.3-005490/1034.9478  !071472..8!48943.-084.3/  \$29 3974/:.80 .07/0503/84390 47/079.7 0..074884.9.7.3.94394.330.-4:990897:.9.9..9:7041 4:7574-02 -:9988349:8:.7030.4:943.09!43974884.4907 .:7390705708039.  -03.07 507.

4:943.9073.70398 0307.  -03.798 .3/  \$29 3974/:.943415439 898420548943.74884.74884.7425:93 0309.8.94394.47928 3 5439.3/42.0754398 \$59.8 .43948054398 :05.07     448037.93-090035.-.

8 94904907 5.74884.0.3/  \$29 3974/:../ 3079.  -03.3.94394.0780./ .4541900301479080.43/.431470.083/0503/03941548943 ..03041901789.47928 &31472.4:943.07     883 0.33./8 944305.7425:93 0309.7039  9.

07 #2:9..94394.4308-09907.  -03.7425:93 0309.4:943.943  0./043/0-.47928 74884.90.3/  \$29 3974/:.

7.088.30.

907/0.90.34907740 2:9.74:3/ 3807 ..0-49 -49.07 43 4:/34947  .943 43 85488-0 4.3 -..0.7002039  9/0503/84390574-02 -:9 30307.. 2.89 7. 9844/94.

7425:93 0309.330.4.0780547.9.  -03.70. 85.08.7039 .7 390.0 9.70.4 4507.70.4:943. 8420070 3-09003 94 5.  0 :83 31472.943 %0708.943.9087.439 / 547.9.943 5928393.078438 9070-89.9438.- :2594.70.47928 74884..839080.3/  \$29 3974/:. :9.70.3331472.9434390574-02 549.073574283.574283.7.3.41 905.3/4282.07 #2:9.8  /.0 92.94394.70.7039 .94380549.0  0 .42509943-09003902  74884.943 .

47928 74884.001706:03.94394. :.74884.078 %49904592:24:41903300/.439 /  3.07/408349.3.  -03.943174294 5. 2:9.943 94:905072039  8431789 -9390545:.3090.3/  \$29 3974/:.943 ..008 74884.33974/:..07 #2:9.943 .03031472.70398 32:9.42-3031472.7425:93 0309.0841 90545:.943 .1907507147233 .74884.07..943    .4:943.943.

3083 90034950.742484208 7..982.4/341390078 89-3..08100.7.:8082.3083905034950 :30 -3...8   390078 4.80714790 4.9438  7.7425:93 0309.94394.4/38.47928 907705708039.3/  \$29 3974/:.553 2.5539.7.0590/9...8980307..-08 ..4/3  \$244907 034950 50349502.935439.998-0990794 03.9./.389.7.2.7..-08/70.920....4/03:207.4:943.  -03...

.039007.47928 39007705708039..-08 0   2.08835.9438     \$420574-0283.7.4:943.  -03.7..0.3/  \$29 3974/:.. .9:7..94394.:081742.0574.209078 9078 9.9047..10/8090   -:0 7003 04 53<  5439.7425:93 0309.

.30147 .9434507.97.7005 8441903:80942:9.-08 4747/3.9047.:31472.:0 #.7005  0 247009424.3/42.4770.074507..4.09482..947847 903/-9 15532:9.7.943942.0 .574-028 98.74884.7/9434.7..0 085 .3/02 .947839..

9431T 3 T :897..0 81:3.:7.94394.  -03.4:943...7425:93 0309.3574-0284.4393:4:85..7.47928 #0.:0/574-028   .870.943 41903:80/3 .943..:0/574-028 0   .3/  \$29 3974/:.2090745928.

2:2570.        c      .943  .742484208 840.  c   Z    (   3 .8434184:943 570.4:943.843 43...4:943 .079-0 4305034950507 034950  <F  (/0130890705708039. .707057080390/ /090723085488-02.47928 .:0843-989738 Z  ( T 7057080390/-.7425:93 0309.  -03. W .55370.3/  \$29 3974/:.94394.:084:94131390.<Z  < W  (F  <2:89-03..

3/  \$29 3974/:.9438       F  @    @  @  @  & ) Z  &314722:9.9354392:9.9354392:9.8.3/42 :31472 1742 & ) 3.44:894-9 1553 -3.7425:93 0309.47928 4.37.0204114.3/427080993 390078  .94394.943 @ /7.9438 0307.4:943.  -03.7 477.

3009..3041.:0 489.94394.8920 .9438  43 :314722:9.7425:93 0309..422432094/894.9439 .:88.90 9.3094.-89.30 ...-0805.3/42/0.0208.  -03.7//0.947.4:943..9354392:9.43 2.82.3/897-:943..-:9:8:.47928 4.. 4898.3/.9094 0.24:39 41.3/  \$29 3974/:..7.7..3/903..:79.30 \$9.//7.031742 9  .439748.0.32094/85745480/ 8:..73 7.70574-..9438 .

.41/0.30 9 94 9 .9438037.

00.-9  47 4:/:803 543947:31472  390720/.70398 03.74884...7..7.   .:0/8  8..7090 0.-0 0  /0503/4390..  -03.93.42-3.07920 ./703 -09003 5.39:31472.70.943  W 5..:034118573 .79209. 070.A  %05.4:943.70398   906:.90 05498/0.947814770.94394.0.3/420.3-0 W .47928 74884. .4389..97.20907.41.7425:93 0309.70.574-.  .3/  \$29 3974/:..074507.79209.0/. A .04190545:.943  ..420817424304198 5..07 W .

830030  ./ 0  9.70398  3 .07801474907./8     .97.3/  3 !.79209.c   .4:943.07 W W W W !.47928 \$30. .c    3 70.74884.3/  \$29 3974/:.3/42  .  .  -03.7425:93 0309..94394.

74884.4:943.3/42030  .19079854392.c  3 70.47928 \$250.c      W  .:08 .70398  3 .7..79209../8    .c  3  .7425:93 0309.c     .07801474907.94394.3/  \$29 3974/:./ 0  9.  -03.3/  3 !.07 W W W !.  .

.  .47928 40.3/  \$29 3974/:.94394.07801474907. c  70.4:943.79209.42243:80/ 3 .3/  !./8 3 .c W  .07 W W W 489.70398  .7425:93 0309.74884./ 0  9.  -03.

47928 !072:9.3/  \$29 3974/:.4:943.7425:93 0309.  -03.943#05708039.9438   7/073.94394.

250%7...0/- .4..347/07 .89413 390078 0..0570880/.:780.943.41.:7309940.. 5072:9.2508479.082.8.479225479.3-084.3574-0281472.77.9438./.943 19070.88 47.0020398 4.:7-0147049078 47/07 .850.39938.03.3384204-0...3!74-02 %\$! 25479.-0890390705708039.700307.7.4907 ...950 %.079...3993 8.983 .00203984.0 .03\$.806:03.  %080574-028.8..703.

7.908    3 W 30.4/3 W .85.943%\$!0.033.943 0  1473    (    (.08 147 .7425:93 0309.  -03..70    5488-094:78 .70  \$0.943705708039.42509094:78430 5072:9.908 W 3/.42509094:79 232.4:943.3/  \$29 3974/:.250    !74-02 W .-090.9089070.94394..039 3.47928 !072:9.

0. 548943 .94394.9079.20907347010.9435.47928 :9.94780.9478..-9 9.09490 408973 7..2:9.33/.  -03..9434507.4:943.7425:93 0309.:04:/20.0.7.9890574-.94309030 .9434507.30..984204507./288-0 84:9438 0  -9 802:9.90./:.9438  472..39.3/ 3443074.8994.94781475072:9.:08 :9..3/  \$29 3974/:.:0 .550/43./943.9  4.:770/   %07014702:89....339484204907.:770/9.30.

7425:93 0309.94.9431475072:9.9438    !.998570807.00.  -03.3/  \$29 3974/:.:08./.90 4909.94394..3/ 90.31472..47928 38079:9..4224/.43/94144901789 819390 7089.943 .03..4394.4:943.97.3/42 4.082489419047/07.09080.

943  38-7403 /87:59847/072470 .08248941.31472.47928 \$.9438   !.008.94.3/8.97.3/42.4:943.7425:93 0309.03..9431475072:9.  -03./.94394.3/  \$29 3974/:.52:9.5907 5489438 !70807.

97.7425:93 0309.31472.  -03.03.04147/07 31472.943 43 -70.9431475072:9.89438 -:9/87:59.47928 3.078432:9.3/42.07990 8:-8973-09003902 !70807./.008.3/  \$29 3974/:.4:943.3/9033.943 .94.082489.94394..9438   !.

97.094-0.8:-809410308.00839480 5489438 34908:-809/408349.9438   !.7.47928 \$.439:4:8 .94394.  -03.3090.4:943.2-02:9.3/  \$29 3974/:..9431475072:9.7425:93 0309..3/42 #.77.3/4270.

  -03,3/  \$29 3974/:.94394;4:943,7425:93 0309.47928 

74884;074507,94781475072:9,9438 

472, .74884;074507,9478419030,/94 3,/288-084:9438 
   

,3850.,80/4507,9478,;0-003/0;80/ .14.:843.42-3347/0747,/,.03. 31472,943174290945,70398

  -03,3/  \$29 3974/:.94394;4:943,7425:93 0309.47928

7/07.74884;07 
 

/0,894570807;070,9;047/079,900203984..:7 31472,574.0/:70  4480,3,7-97,75,7917429017895,7039  45985,7994901789./  45903:2-0789,9,7034939017895,79 94 901789./  89,793791742.:954394190.450/5,79   :839047/07 419080.43/5,7039  ,3/7,553,74:3/,99003/  3,44:81479080.43/./ 95,70397408 70;0780/

  -03,3/  \$29 3974/:.94394;4:943,7425:93 0309.47928

7/07.74884;070,250  

457,3/42800.90/809174217895,7039  

457089174280.43/5,7039347/07    

3-010/1742!  \$0..:50/- 3!  ..450/3985.47928 !.8..7931742901789.450/  470.419080 44390411857394800.3/.3/42802039.3/  \$29 3974/:.7425:93 0309.44:8 ./34118573  1905..550/74884.07 ! 31472.  -03.00349.07802039 907089 41904118573.04.90 349-05:993 9070 .:50/- 3!.8 -003..70398!.3.90/.88.07543944147002039839.43/.74884.0/:701475.70.3/!  44807../-00310/390 4118573 5:9 3905489434.99900020398174290.0349-003.574.94394.9 80203941!9..01742!  !../8.74884.4591742!  \$9.9002039 .70.9..4:943.0 394905489434.79.70.3/0.:50/ 3! 83...

7425:93 0309.  -03.4:943.3/  \$29 3974/:.250  \$905  \$905  \$905 .94394.47928 !0.

74884...039017895. \$9.03901789.4:943.0174280.. 494905489439908.98905-974:/:394:..3/  \$29 3974/:.0 0 #050.0.  -03.94394.990.041.20548943 3! .7039940907998548943 31472.47928 .7425:93 0309.0309.0.20.0084190.  .420817424305.9908.799901789. .0081742!3901443.0/:70  ../43905489438 90.7039 .44.43/5.07 .00.574.0041!  ..00 3!  / //98../0.7039  %.8.009490.00.9901789..0.0041!  !:990.77.

0.90.45.083944118573 .9073..070.47928 .08  \$905.74884.7425:93 0309.250  \$905/0391..94394.  -03.3/  \$29 3974/:.4:943.

943   478-.705708039390945.94394.3 0/08.  -03.-0893.43897:.7425:93 0309. 0  (.79.9.4224394-49 2.47928 /0#0.3/  \$29 3974/:.4:943. 0/08.70398 1.3/ ( .93.42-3.

:770390020390397  #024.574.943 31472.94394.23090490703/4190411857381470903843 90780.:77039002039 190708.23089147..47928 /0#0.422430/0 5..  -03.4:943..43897:.4803.0.002039.3/  \$29 3974/:.300020398.7.7425:93 0309.00/09..3/5:993904118573  \$0990.994-0309002039 907805.70859..0/:7043.-0  .3.97.339.0894.890847908989 %08.804170.3025989 .9.3/42.97.90/  !.3/42 .-08.97.:770390020391742909.-0.42-3.7010703.9801.3/42  390.90039739089.

42-3.9430.250 .  -03.3/  \$29 3974/:.94394.7425:93 0309.4:943.47928 /0#0.

9.08 0  5 80:..80/433:207.9478 ..94384370.908 413.4930307.4507..:0/.42-3.3/  \$29 3974/:.9.0 8 897.001706:03.4:943.83.703970.74884.79209.940.39508 .07 ...74884.230 003.74884.07 0307.94394.008 0   .070307.74884.943:8085.83.7039 .47928 :95.90/-9057.943     #0. .83 :31472..74:3/83.833 5439.  -03.43.3/ 97.9434507.03907412.90.70349.42-3.07 90090384394.70-:989:/083/.74884.88.7425:93 0309.43897.70398 0   /..9:70 4939.80/43802039./943.943.80/43.07 .42-3.92:9.94341905.3/70. 70.  .90 :801: %7002.9:7..

943705.943 430202-0741545:.70\$90.0/  147 .3/.47928 !45:.8:7.90/5070307..90 24/08 430411857380307.9434/08  \$:808.0/  0307.3/  \$29 3974/:.94300307.7425:93 0309.943.0/-904118573  990490703/41908.4:943.  -03.24/0 0.081470.0307.0./:.943705..5 905745479434190545:.943....94394.943 9003970809415...703988705./ \$9.

545*80 147\$\$ .

47928/01304574-.943949.943 8:7..943   \$00..:770390307.  0 90.08 \$00.05..9431742.70398 4118573944394309 0307.70705708039.94317425.80/42509943  \$00.943  \$00.793 2.943 3/0503/039 893.  -03.3/  \$29 3974/:.943-09003 800./:.47928 93088.:73945.93 5.943 4507..47800.943.7039800.7425:93 0309.9434507.943574-.34.94394.-908 .-908.9478/0130800.7025020390/ .94784743403/.4:943.

4:943.250\$  50.90/3:2-0741.47928 2502039./:.450841.  3 3 W 1  .94394.9430.  -03.3/  \$29 3974/:.33/.7425:93 0309.

390081447  3  A 3 A .:0413 3 545 80 1  19308841  1 ..0  3  .4792 3 0.03.07 8\$&\$.9438 4:.7.574-.0 :.3/85343.3900843.0385.0/.. 1930883545   .720/ 003 9208942.4792 . 1 #4:099000.-9/897-:943 853.9:.7284300.03 800.. ...7.

943  !74-0283.47928 93088 !74547943..90\$00.:/0 3019202-07. 84789193088398 .4:943.059-0941:3. .94397.9:7043.071708941 545:.5/9.3. 1    1 . W 91  070.39 :8:.31  2..9438 \$2.31.7425:93 0309.38548943  \$.7 480 800.37.94357088:70 8:8..  -03.\$.94394.0 903/417:380319308808.8994574-028 3/431  1   ...08819!702. 9  070 .94382:.7082.04.0703. 8.4389.3/  \$29 3974/:.893 0307.

3/  \$29 3974/:.38548943147!\$ .4:943.94394.47928 :3.94397.  -03.7425:93 0309.

80/\$00.3 %8254808.87.0574-02841!\$--.9079.30-0 .070.-84:90 193088 #.4:943.7425:93 0309.3/903 -.943    990259947024.4792 -:9988:8:.3545:.  -03.943920 .3/  \$29 3974/:..-9084370..3070 199089.847934..47928 #.3 ./4390 .0 7.70/94901930880.943574-.47/394193088.94394.80800.943574-.83 800.3/47897.3 .33 .9.:.425.-908437.943.

3/  \$29 3974/:.8:708.041-0893/.2090780/-1.94394..4:943.  -03. 3988903:2-0741.7#.250 .9478 8 A  20.39./703../:./.4990/949 \$250202-070.33   !.47928 30.7.7425:93 0309.

#.3/  \$29 3974/:.80.391.338290/94800.4:943.33.94380 ..3 ./:.947. 472.7425:93 0309.47928 543039..  -03.7#..94357088:70 543039..4508941990893/.4.3.33    30.94394.#.47/394 545:.4389.. .9024709.

04109073.545:.4.7425:93 0309.308 #00843570803..943 .-499030.-4.94394.3/  \$29 3974/:.994800.990-08941 9080 #050.293490890  0.205.3/42903800.078  31472.924703/.!74.02.97.3./:.1930881:3.8 4:/-0.989.0/:70 !.943  2094/8.085 435.2039\$00.7.  -03.4:943.47928 %4:73.070434-. 202-078.943 89.8 .

550389574-.70.4:943.  -03.439089.2039\$00.808800.439089..39.02039  0907199089. 4798.020393.394:9705.398.-941800.94357088:70 0907.705.830.0/9705.838 /09072389.250  07 3.04.949.77.808800.93 /0503/43 #.-95  47  9201471990893/./:..20.94357088:70 !.07 545:.3398W 5 .94394.7425:93 0309.70.341 \$0418.3/  \$29 3974/:.943  !74-.47928 %4:73.9438908..

3-0/.943  .. /0090 4/089  &8343041902094/8.3/42  349 70.  -03.32502039..0:80/1475.8 /0090 7.42203/0/ 47...3/  \$29 3974/:.943.  .7039 800.943   489412094/8.94394.08 0 ./0/39494 .4:943.47\$00.47928 \$:7.81789 3 1789 4:9 .5574.-4.047 93088 .80/\$00.47800.-4.943   0  \$ 3\$\$.7425:93 0309.943 \$:7.80/\$00..

5/9.3/  \$29 3974/:.94324/08   \$\$ .943847 34 /:5.90.7  % #.94394..  -03.331475.77.07:809. ./ \$9.70545:.90.943 #..47928 %4\$50. /0090 4789 74290 8473.4792 0 .4:943.7425:93 0309..808  982 /0:80/3-49545:.89430.84:80/30.8005.  .45419019908984:943 841.\$90.04.7039800.908 54.

43/9438/0130/--0 .94394.94341908.47928 .35071472.8..7425:93 0309.0/:08232.55.9.0/0..943414Z 432Z  %04.9438 !70  L /01308.0/:09.0/03.809414507.43897. .80/8\$\$! !70.9810/  592.250.809412. :7  L F #/0130890/:7. 4507.30/4-8458..4507.4-8.9438 8.308.3/  \$29 3974/:.  -03..3/!70 .708.0/:0/  4770./:7.9 8  425090..308 -0  L /01308.94384:/570.8349413/.708.9434147/07-.0.809414-8 8.0/:3574-02       8.90949.4:943.2.

0901789 309 4507..79392043982.70..5072:9.943.943433.7      19308841.0780193088 0307.0/:3   #05708039.7425:93 0309.88:209070843430 .90/841.:5.74884.070.943 :807.943 570..24/01478:7.304.994   2. 44:5982.0.4/3574.9438414507.8.943 .30/4-8458.0/:08-.705072:9.94394.9438 !072:9..47708543/3 8.8.943.3/.70895488-089.9438.4:943.3/4239.943890/:7.30 8:-0.70/0.883900.43897.9434190.30 0700.9433908.7039800.94384/3147984507.-02:9.47928 !70.  -03..0/03.0/03.4/0/948.07 :8074:0990005./:./:.38:9.47800.9433/.0/:0./0.3/  \$29 3974/:.9431742903/.0/:70 9.0/:0 94-02320/ :80.

425.47928 \$\$!0.  -03.94394.947.3/  \$29 3974/:.7425:93 0309.7843 .4:943.2504507.