You are on page 1of 5

Chaos Updating Rotated Gates Quantum-inspired

Genetic Algorithm
Chen Hui, Zhang Jiashu, Zhang Chao
! . Signal and Information Processing Key Laboratory of Sichuan Province,
Southwest Jiaotong University
Chengdu, 610031, China
Email: kithcx@sohu.com

Absrrod--From the recent research on combinational Quantum Calculation Theory was presented in recent
optimization of knapsack problem, quantum-inspired genetic years. As for its ability to do parallel calculations,
algorithm (QGA) was proved to be better than conventional Quantum Evolutionary Algorithm has become a research
genetic algorithms. To accelerate the convergence speed of the hotspot [‘I. For example, quantum genetic algorithm (QGA)
QGA, this paper proposes research issues on QGA such as has drawn research emphasis in many countries. These
Q-gate. A novel Q-gate updating algorithm called Chaos kinds of algorithms do evolutionary operabion and storage
Updating Rotated Gates Quantum-inspired Genetic based on quantum calculation theory. But as for the
Algorithm (CQGA) is proposed. Analysis of the two main quantum calculation still has many problems to be settled,
characters of quantum computing and chaos was also the method has great difficulty to be realized. Paper [ I ]
presented. This algorithm gives the convergence proposed quantum GA based on Von Neumann computer.
demonstration of the Quantum Genetic Algorithm This method was inspired by the quantum bit definition
(QGA).Then several experiments are carried out on a class of and hidden parallel process. The improved GA has
numerical and combinatorial optimization problems. The advantages on the result variety and colony scale. As for
results show the updated QGA make QGA more powerful the difficulty in duplicating a better quantum bit for its
than the previous QGA in convergence speed. storage and running time, the QGA only take quantum
Keywords: chaotic quantum genetic algorithm, quantum mutation into account. But there is still discussion on
genetic algorithm, quantum computing, chaos whether the method is simple and efficient. Considering
that the chaos sequence is addible and the quantum bit in
;he evolutionary algorithm is chaotic, a more efficient
I. INTRODUCTION Chaos Updating Quantum Rotated Gates Algorithm is
Genetic algorithms (GAS) are principally a stochastic proposed. The method is applied into QGA to fonn CQGA
search and optimization niethod based on the principles of with good simulation results.
natural biological evolution. That is, Genetic Algorithm is
an inspired iterative calculation method firstly proposed by
professor Holland of Michigan University in USA. The
11. QUANTUM CACULATIONAND CHAOS
G A does the guided search instead of random by DISTRIBUTION OF QUBIT
simulating the evolutionary procedure of the natural Feynman firstly brought forward the Quantum
creatures. It is very useful to deal with the nonlinear Calculation theory in 1982, and Deutsch then pointed out
reversal problems in the manual work systems to search that the quantum state of entanglement could be used to
for the cotnplex special result [’I, De Jone firstly put the realize parallel calculation. Then the quantum calculation
GA into use foF focusing on the function optimization became hot [71. Basic characters involved in the quantum
problem; his research work shows that the GA is an state are the quantum states superposition, interference,
efficient method in dealing with the Mathematical entanglement and nonclonability [71, In the quantum
Programming. It doesn’t put limitation to the formats of calculation, information cells are called qubit, which could
model structures, optimal rule formats, and the amount of have superposed state other than simple states “0” or “1”.
the parameters to be estimated and circuinstance whether And the superposed states are the random linear addition
there are some restriction conditions. It suggests great of “0” and “1”. They could be either “0” or “1” just
application foreground in parameters estimation, pattern decided by the probability of the coexistence [8’. In this
recognition, machine learning, neural networks and case, n qubits infomiation could express the quantity ,of
industry control. But as to huge and complex systems, information 2“ times of the traditional ones.
especially to the nonlinear systems, the traditional GA also Though quantum calculation has advantages of parallel
has shortcomings such as huge calculating storage, losing process and less storage, it has many difficulties in
of the best chromosomes, less guarantee for converging to realizing [‘I. Here we won’t talk about the physics reality
the whole optimal result earlier of the quantum calculation, but put our emphasis on

0-7803-8647-7/04/520.00Q 2004 IEEE. 1108


analysis of the quantum-inspired GA. T=t+ 1;
In the recent research, Andrew Noble Jordan discovers Generate P(t) from Q (t-I)
that the qubits have the characters of chaos. That is to say,
Evaluating P (t)
when qubits express the result space, each qubit has the
probability of chaos disuibution appearing “1” or “O”Lzl. Updating Q(t) using rotated gates look up table
This character has great significance to cognition the Save the best of P (t)
updating mechanism of qubits and reduction the cost of the End
quantum rotated gates calculation.
The proposal of his algorithm is mainly based on the End
following chancters:
Lemma 1: The random combination of two chaotic time
sequences is still a chaotic sequenceLP1.
Lemma 2: The qubit of the quantum calculation is IV. CHAOTICQUANTUMGENETIC ALGORITHM
chaotic [’I, AND ITS CONVERGENCE
Deduction 1: The qubit in the quantum evolution A . Chaotic quantum genetic algorithm
calculation still has characters of chaos after passing the Chaos is a universal phenomenon in the nature, which
rotated gates updating.
Illustration: It is known that the qubit is chaotic seems like ball-up but have exquisite inherent structure. It
according to the lenima 2, and the quantum evolution has characters of “random”, “ergodicity”, “rule”, which
begins with initializing qubits into the same constant. Then could experience all over the states in a certainly range
after perpetual updating has rotated gates updating the
qubits, the final qubits should have the same characters as according to its own rules. Chaos had been put into good
the qubits without obligation. That is to say, after the use in optimal problems such as random search to gain
rotated gates updating operation, the qubits still have the improvement of the traditional optimal algorithms 16’ ’I.
characters of chaotic distribution.
As for the chaos character of qubit, the seed groups of
QGA after the updating quantum gates still have the
STRUCTURE OF QGA
111. FAMILIAR
QGA is an evolution algorithm based on the quantum character of chaos. As for the loop-up table operation has
calculation theory, which adopts quantum coding to many compare and calculation, the complex is very high.
express chromosome. As for the probability expression of We proposed the chaos updating quantum gates for
the qubit, it could express the linear entanglement of the
results. So it has better ability to search for the optimal mutation algorithm based on the same distribution of the
result than the traditional GA1” qubit as chaos, which is described:
A qubit could be either%“ or “l”, then it could be
described: Step]: initializing
Initializing size of the seed group n, max genetic
(Y)=a(O)+P\l) (1)
number N, bit number of the chromosome in and
evaluation function f (x), current genetic t, quantum
Hereinto, a , B are complex numbers, which denote
rotated gates, seed group Q(t) .
the corresponding state appearance probability, and they
Step2: generate chaotic sequence
meet the restriction condition:
Here we get chaotic sequence and super chaotic
sequence using these two equations:
la\2i / P l 2 = 1 (2)
Thus it can be seen, n qubits could denote 2” states at y(t + 1) = 4y(t)(l- y(t)) , y (1) #0.5 and
the same time.
The basic structure of QGA is described: o<y ( 1 ) < 1 (3)

Ix(t + 1) = -O.lx(t) +k(t)


Procedure QGA
Begin \k(t+l)=x(f)’ -1.6
Initialize cunent generation t=O
Initialize Q (t) l e t x (1) =y ( 1 ) =0.3 (4)
Compare Q(t) with random r ( r E (0, 1 ) ) b i t by bit, gets Step3: Compare Q(t) with random r ( r E ( 0 , 1 ) ) b i t by
P(t) bit, gets P(t)
Evaluating P (t) Evaluating P (t)
Reserve the best of the P (t) Reserve the best of the P (t)
While (termination condition) While (termination condition)

1109
~

t=t+l; the generation number as 50 and the seeds number 50;we


Generate P(t) from Q (t-1) also set different precision for different test functions.
Evaluating P (t) Firstly we make a survey over a multiple peaks function
Updating Q(t) by using chaotic sequence updating optimal problem:
quantum rotated gates.
Save the best y=x+lOsin(5x)+7cos(4x) (5)
End
End

Equivalence stated as following:


A s for the ergodicity of chaotic sequence, the chaotic
sequence could increase the multiformity of the seed group
after mutation. Conclusion can be drawn from the lemma 1
and lemma 2 that it will not change the qubit chaotic
distribution using a direct chaotic sequence to updating
quantum rotated gates instead of using look-up table.
Fortunately it will save the time of looking up and
comparing, and the optimal algorithms could execute
efficiently.
B. The convergence of' chaotic quantum genetic
algorithm
The convergence proving of quantum evolutionaty
strategy states below " I :
Definition 1: strictly overlay mutation distribution
Suppose the random vector of Z E %" , S=diag ( s , .
s l , ..., sn) is a full order on the cross matrix, which meets
/SI=l si 3 smin > 0 , T is arbitrary positive orthogonal
I

matrix, (I is the variable in the positive real number subset,


namely strategy variable. If mutation variable
AZ = oTSZ,we take the mutation distribution as strict
ergodicity.
Lemma 3: As to boundary non-protruding function,
considering it has inhibited of non-zero measurement
limitation. If the evolutionary algorithms meet the
following conditions we could call it convergence [8]:
0: Adopt evolutionary selecting
0:Adopt replacement for recombination
0:Mutation distribution keeps strictly ergodicity.
Deduction 2 : chaos updating rotated gates quantum
GA is of whole convergence.
Illustration: because
a d u e 'to lemmaland deduction 1, chaotic quantum
GA has equivalence with basic CA.
@quantum GA adopts evolutionary selecting
@quantum GA adopts "full disturb across" mutation
operator to replacement and recombination.
@ all transfonn matrixes are all quadrature in
quantum mutation, the distribution of this mutation is
strictly laid over, that is to say, the chaotic quantum
mutation is strict ergodicity. Fig.e3 Basic QGA convergence curve
Thus, the Chaotic Quantum Genetic Algorithm is full
convergence.

V. SIMULATION
In the interest of doing contrast investigation, we set all

1110
F 2 = [ 1 + ( x l + x 2 +1)2(19-14x, +3x, 2 - 1 4 ~+ 6~x l x , +3x2’)1

* [30+(2xl - 3 ~ ~ *) ’

(18-32x, + 1 2 x 1 2 + 4 8 x 2 - 3 6 x l x 2 + 2 7 ~ , ~ ) ] ,

Here - 2 5 x i 5 2 (7)

sin2Jm-0.5
F 3 = 0.5 - ,Here
Fig.4 SCQGA convergence curve (I + 0.001(x,2 + x2 ))
-100 5 xi 5 100 (8)
Tablel: The Stat. of %vera1 optimize algorithm execution
100 times each
* { ~i = il c o s [ ( i + l ) y + i ]
I Algorithm I T C I CAV I GN I CR
+ O.S[(.r + 1.42513)2 + ( y + 0.80032)’] (9)
CQGA 0.48s 24.8552 42 93%
I SCQGA 10.73s I 24.8496 I37 I 85% I where FI,F2, F4 for least, F3 for maximum.
Conclusions can be drawn from the experiments that the
basic QGA cost much more time than the improved
Note: TC=time cost; CAV= convergence average value algorithm. As for the reason that the quantum updating
GN=generating number CR=convergence rate rotated gates are decided by looking up tables which need
To test the efficiency and general fitness of all the multiple parameters judgment tree to search. What’s more,
optimize problems, we set four general functions to do the every qubit need its own rotated gates to updating the
test, each function was optimized separately.100 times for quantum gates, so its time complex would be great. On the
each method. The Stat was listed below as table 2: foundation of analysis of equivalence of the two
algorithms, we proposed novel method of having a limited
Table2 CQGA and basic QGA convergence and efficiency scope chaos sequence directly updating the q to reduce the
time cost of looking up tables and comparing. It makes the
TF OPC FE EG BGA- ACT(s) GA more efficient. It provides foundation to apply the
QGA in the real time environment.
I I I R I EGR I This I Fomi I Thus it can be seen, simply using chaotic sequence to
update chaotic quantum gates is a feasible method. It fully
F1 I (I,]) 1 0 83 85 11.53 1.06 takes quantum state chaotic characters and chaotic
sequence ergodicity together into good use to improve the
F2 I (0,-I) I 3 I75 78 I 11.68 I 1.13 traditional quantum-inspired GA by structure and
F3 I (0.0) I 1 193 I 90 I 2.03 I 0.49 efficiency. Other attempts such as using super chaos
F4 (-1.425 -186. 76 75 23.08 2.59 sequence shows that the better one is chaos.
13,-0.8 7307
VI. CONCLUSIONS
0032) Aiming at the disadvantages of quantum-inspired
genetic algorithm, chaotic sequence updating quantum
Note: TF=test function OPC=optimal point coordinates rotated gates method is proposed and testified. And the
FE=full extremum EGR=extremun gotten rate analysis of the algorithm also shows great potential
BGA-EGR=basic GA extremum gotten rate application in optimal fields. The farther work will be
All the test functions are listed: focused on the mechanism of the chaotic inspired quantum
genetic algorithms and makes it perfonning better.
Fl =100(x,2 - x 2 ) 2 + ( 1 - x t ) ? Here,
ACKNOWLEDGEMENT
- 2.048 5 xi 5 2.048 (6) This study was supported by the Southwest Jiaotong
University Fundamental Research Funds (No. 2001 BOX)

1111
and the Scihuan Province Outstanding Youth Funds (No.
0324026-033) .

REFERENCES
[ I ] Bar Rylander, ”Computational Complexity and the
Genetic Algorithm”,PHD paper, College of Graduate
studies University of Idaho,2001.
[2] Andrew Noble Jordan, “Topics in Quantum
Chaos”,PHD paper, university of California Santa
Barbara. 2002.
[3] Kuk-hyun Han, Kui-hong Park, “Parallel
Quantum-inspired Genetic Algorithm for
Combinatorial Optimization Problem”, I€EE,
Evoltitionary Computation. V01.2, pp. 1422 - 1429
May2001.
[4] Ajit Narayanan, “An introductoly tutorial to
quantumConiputing, Quantum Computing”, Theoy,
Applications and 1mpl;cations. IEE Colloquium on, 4
June 1997.
[SI Xi Yugeng . Chai Tianyou, “survey of genetic
algorithm”, control theory and application, Vol. 13,
N0.6, ,pp.687-708. 1996
[ 6 ] Li Chuangfeng, Guo Guangcan, “progress in quantum
information research”, Progress in physics, V01.20,
No.4,pp.407-43 1,2000.
[7] Guo Guangcan, “introduction, to quantum
information”,Phj~sics China , Vo1.30 ,. No.5 ,
pp.286-293.2001,
[8] Yang Shuyuan, Liu Fang and Jiao Licheng, “the
Quantum evolutionary strategies”, ACTA
ELECTRONICA SINICA , Vo1.29 No.l2A ,
pp.1873-1878. 2001.
[9] Gan Jianchao, Xiao Xianci, “Characteristic of addition
of chaos”, ACTA PHYSICA SINICA , Vo1.52,No.5,
pp.1085-1090,2003.

1112

You might also like