Transcript Computational Complexity - 서울대 : Biointelligence lab

Computational Complexity
Jang, HaYoung
([email protected])
BioIntelligence Lab.
Algorithm Analysis
Why Analysis?
to predict the resources that the algorithm
requires, such as computational time, memory,
communication bandwidth, or logic gates.
The running time of an algorithm is the
number of primitives operations or
“step”(machine independent) executed.
Complextity
Space / Memory
Time
Count a particular operation
Count number of steps
Asymptotic complexity
Time Complexity
Worst-case
an upper bound on the running time for any
input
Average-case
We shall assume that all inputs of a given size
are equally likely.
Best-case
to get the lower bound
Time Complexity
Sequential search in a list of size n
worst-case : n times
best-case : 1 times
n
1
( n  1)
average-case :
i

n i 1
2
Asymptotic Notation
Asymptotic upper bound, we use notation.
For a given function g(n), we denote by 
(g(n)) the set of functions; (g(n)) = {f(n):
there exist positive constants c and n0 such
that f(n) <= cg(n) for all n >= n0}.
Asymptotic Notation
-notation provides an asymptotic lower
bound.
For a given function g(n), we denote by (g(n))
the set of functions (g(n)) = {f(n): there
exist positive constants c and n0 such that
f(n) >= cg(n) for all n >= n0}.
Asymptotic Notation
(g(n)) = {f(n) : there exist positive
constants c1, c2, and n0 such that c1g(n)
<= f(n) <= c2g(n) for all n >= n0}.
Asymptotic Notation
The sets (n2), (n2), and (n2)
Practical Complexities
109 instructions/second computer
2
3
n
n
nlogn n
n
1000
1mic
10mic
1milli
1sec
10000
10mic
130mic
100milli
17min
106
1milli
20milli
17min
32years
Impractical Complexities
109 instructions/second computer
4
10
n
n
n
1000
17min
3.2 x 10
years
10000
116
days
10^6
3 x 10^7 ??????
years
???
n
2
13
3.2 x 10
years
???
??????
283
Faster Computer Vs Better
Algorithm
Algorithmic improvement more useful
than hardware improvement.
E.g. 2n to n3
Intractability
•
A polynomial-time algorithm is one whose
worst- case time complexity is bounded above by
a polynomial function of its input size.
W(n)  (p(n))
•
example - worst-case time complexity
- Polynomial-time : 2n, 3n3 + 4n, 5n + n10, n log n
n
0.01n
n
- Non polynomial-time : 2 , 2 , 2 , n!
•
Intractable problem - No polynomial-time algorithm can
solve it.
Three Categories of Problems (1)
1. Problems for which polynomial-time algorithms
have been found
2. Problems that have been proven to be intractable
- The first type is problems that require a nonpolynomial amount of output.
e.g. Determining all Hamiltonian Circuits.
- The second type of intractability occurs when our
requests are reasonable and we can prove that the
problem cannot be solved in polynomial time.
e.g. Halting Problem, Presburger Arithmetic Problem
Three Categories of Problems (2)
Presburger Arithmetic is the theory of integers with addition
(Z,+,=,<,0,1) and is known to require doubly exponential
nondeterministic time.
3. Problems that have not been proven to be
intractable but for which polynomial-time
algorithms have never been found
e.g. 3-SAT, 0-1 Knapsack, TSP, Sum-of-Subset, Partition,
Graph-Coloring, Independent Set, Vertex-Cover,
Clique,3D-Matching, Set Cover, etc.
The Sets P and NP (1)
•
Definition
P is the set of all decision problems that can be
solved in polynomial-time.
•
A NP algorithm have two stages:
1. Guessing (in nondeterministic polynomial time) Stage
•
2. Verification (in deterministic polynomial time) Stage
Definition
NP is the set of all decision problems that can be
solved in nondeterministic polynomial-time.
The Sets P and NP (2)
Genetic Algorithms
An Abstract View of GA
generate initial population G(0);
evaluate G(0);
t := 0 ;
repeat
t := t + 1;
generate G(t) using G(t-1);
evaluate G(t);
until termination condition has reached;
Search Techniques
SEARCH TECHNIQUES
Calculus-based
techniques.
Directed
Methods
Fibonacci
Newton
Guided Random
Search Techniques
Indirected
Methods
Simulated
Annealing
Enumerative
Techniques
Evolutionary
Algorithms
Evolutionary
strategies
Dynamic
PGMing
Genetic Algorithms
Parallel
GAs
Classes of Search techniques
Sequential
GAs
Simple Genetic Algo's
components
1. A mechanism to encode the solutions
as binary strings
2. A population of binary strings
3. A fitness function
4. Genetic operators
5. Selection mechanism
6. Control parameters
The GA Cycle
Population
(chromosomes)
Offspring*
Crossover &
Mutation
New Generation
Genetic
Operators
Parents*
Manipulation
Mates
Selection
(mating pool )
Decoded
strings
Evaluation
(fitness)
Reproduction
= Evaluation
+ Selection
Fitness function (object function)
The mechanism for evaluating each string
To maintain uniformity over various
problem domains, normalize the
obj.function to 0 to 1.
The normalized value of the obj. function
=
the fitness of the string
Selection
Models nature's "survival-of-the-fittest "
mechanism
A fitter string receives higher number of
offspring.
Proportionate selection scheme
The roulette wheel selection scheme
Crossover
After Selection, pairs of string are picked
at random
If string length = n,
randomly choose a number from 1 to n 1,
then use it as a crossover point.
GA invokes crossover
ONLY IF a randomly generated no > pc .
(pc = the crossover rate)
Mutation
After crossover,
string are subjected to mutation
Flipping bits : 0
1,
1
0
Mutation rate : Pm
= probability that a bit will be flipped
The bits in a string are independently
mutated.
= Role : restoring lost genetic material
Function Definition
Find x from the range [-1, 2] which maximizes the f
f ( x)  x  sin(10  x)  1.0
Analysis of function f
f ( x)  x  sin( 10π  x)  1.0
f ( x)  sin( 10π  x)  10πx  cos(10π  x)  0
tan(10π  x)  10π x
2i  1
 ε i , for i  1, 2, 
20
x0  0
xi 
xi 
2i  1
 ε i , for i  1,  2, 
20
Representation (1)
Representation of string
six places after decimal point
The range [-1,2] should be divided into at least 3000000
ranges.
Binary representation : 22 bits
2097152  2 21  3000000  2 22  4194304
Mapping from a binary string
a real number x
b21b20 b0
into
Convert the binary string from the base 2 to base 10:
b b
21
20 b0
  
21
2
i
b

2
i
i 0

10
 x
Find a corresponding real number x:
3
x  1.0  x  22
2 1
Representation (2)
String Example :
string (1000101110110101000111)
x  (1000101110110101000111)2  2288967
x  1.0  2288967 
3
 0.637197
4194303
String
0000000000000000000000
1111111111111111111111
x
-1.0
2.0
Initial Population
Create a population of strings, where each
chromosome is a binary vector of 22 bits.
All 22 bits for each string are initialized
randomly.
Evaluation Function
eval(v) = f(x)
For example,
Strings
x
f(x)
v1=(1000101110110101000111)
v2=(0000001110000000010000)
v3=(1110000000111111000101)
x1= 0.637197
x2= -0.958973
x3= 1.627888
f(x1)= 1.586345
f(x2)= 0.078878
f(x3)= 2.250650
The string v3 is the best of the three strings,
since its evaluation returns the highest value.
Genetic Operators : Mutation
Mutation
Mutation alters one or more genes with a
probability equal to mutation rate.
Mutation Example :
v3=(1110000000111111000101)
Fifth gene
v3’=(1110100000111111000101)
x3  1.721638
f ( x3 )  0.082257
10th gene
v3’’=(1110000000011111000101)
x3  1.630818
f ( x3)  2.343555
Genetic Operators : Crossover
Crossover Example :
The crossover on v2 and v3
Assume that the crossover point was randomly
selected after 5th gene:
Before Crossover
v2=(00000 | 01110000000010000)
v3=(11100 | 00000111111000101)
x2= -0.958973
x3= 1.627888
f(x1)= 0.637197
f(x2)= -0.958973
After Crossover
v2’=(00000 | 00000111111000101)
v3”=(11100 | 01110000000010000)
x2’= -0.998113
x3”= 1.666028
f(x1’)= 0.940865
f(x2”)= 2.459245
Parameters
Population size = 50
Probability of crossover = 0.25
Probability of mutation = 0.01
Experimental results
The best chromosome after 150 generations
Vmax = (1111001101000100000101)
Xmax = 1.850773
xmax  1.85  ε
As expected,
, and f(xmax) is slightly
larger than 2.85.