Improved Harmony Search Algorithm with Chaos for Absolute Value Equation

In this paper, an improved harmony search with chaos (HSCH) is presented for solving NP-hard absolute value equation (AVE) Ax |x| = b, where A is an arbitrary square matrix whose singular values exceed one. The simulation results in solving some given AVE problems demonstrate that the HSCH algorithm is valid and outperforms the classical HS algorithm (HS) and HS algorithm with differential mutation operator (HSDE).


Introduction
We consider the absolute value equation (AVE): where , and x denotes the vector with absolute values of each component of x .A slightly more general form of the AVE (1) was introduced in John [1] and investigated in a more general context in Mangasarian [2].As were shown in Cottle et al. [3] the general NP-hard linear complementarity problem (LCP) that subsumes many mathematical programming problems can be formulated as an absolute value equation such as (1).This implies that AVE (1) is NP-hard in general form.Theoretical analysis focuses on the theorem of alternatives, various equivalent reformulations, and the existence and nonexistence of solutions.John [1] provided a theorem of the alternatives for a more general form of AVE (1), Ax B x b   , and enlightens the relation between the AVE (1) and the interval matrix.The AVE (1) is shown to be equivalent to the bilinear program, the generalized LCP, and the standard LCP if 1 is not an eigenvalue of A by Mangasarian [4].
Based on the LCP reformulation, sufficient conditions for the existence and nonexistence of solutions are given.
Prokopyev proved that the AVE (1) can be equivalently reformulated as a standard LCP without any assumption [5] on A and B , and discussed unique solvability of AVE (1).Hu and Huang reformulated a system of absolute value equation as a standard linear complementarity problem without any assumption [6] and gave some existence and convexity results for the solution set of the AVE (1).
It is worth mentioning that any LCP can be reduced to the AVE (1), which owns a very special and simple structure.Hence how to solve the AVE (1) directly attracts much attention.Based on a new reformulation of the AVE (1) as the minimization of a parameter-free piecewise linear concave minimization problem on a polyhedral set, Mangasarian proposed a finite computational algorithm that is solved by a finite succession of linear programs [7].In the recent interesting paper of Mangasarian, a semismooth Newton method is proposed for solving the AVE (1), which largely shortens the computation time than the succession of linear programs (SLP) method [8].It shows that the semismooth Newton iterates are well defined and bounded when the singular values of A exceed 1.However, the global linear convergence of the method is only guaranteed under more stringent condition than the singular values of A exceed 1.
Mangasarian formulated the NP-hard n-dimensional knapsack feasibility problem as an equivalent AVE (1) in an n-dimensional noninteger real variable space [9] and proposed a finite succession of linear programs for solving the AVE (1).
A generalized Newton method, which has global and finite convergence, was proposed for the AVE by Zhang et al.The method utilizes both the semismooth and the smoothing Newton steps, in which the semismooth Newton step guarantees the finite convergence and the smoothing Newton step contributes to the global convergence [10].A smoothing Newton algorithm to solve the AVE (1) was presented by Louis Caccetta.The algorithm was proved to be globally convergent [11] and the convergence rate was quadratic under the condition that the singular values of A exceed 1.This condition was weaker than the one used in Mangasarian.
In this paper, we present an improved harmony search with chaos (HSCH).By following chaotic ergodic orbits, we embed chaos in the pitch adjustment operation of HS with certain probability (RGR).Moreover, chaos is incorporated into HS to construct a chaotic HS, where the parallel population-based evolutionary searching ability of HS and chaotic searching behavior are reasonably combined.The new algorithm proposed in this paper, called HSCH.Simulation results and comparisons demonstrate the effectiveness and efficiency of the proposed HSCH.
In section 2, we give some lemmas that ensure the solution to AVE (1), and present HSCH method.Numerical simulations and comparisons are provided in Section 3 by solving some given AVE problems with singular values of A exceeding 1. Section 4 concludes the paper.
We now describe our notation.All vectors will be column vectors unless transposed to a row vector.The scalar (inner) product of two vectors x and y in the n-dimensional real space n R will be denoted by , x y .The notation will signify a real m n  matrix.For such a matrix T A will denote the transpose of A .For the 2-norm will be denoted by ||A||.

Research Method 2.1. Preliminaries
Firstly, we give some lemmas of AVE, which indicated that the solution to AVE is unique.The following results by Mangasarian and Meyer [4] characterize solvability of AVE.
Lemma 2.1 For a matrix , the following conditions are equivalent.
(i) The singular values of A exceed 1.
(ii) (ii) The AVE (1) is uniquely solvable for any It is clear that * x is a solution of the AVE (

Classical Harmony Search Algorithm
Since harmony search (HS) was proposed by Geem ZW et al. [16], it has developed rapidly and has shown significant potential in solving various difficult problems [17].
Similar to the GA and particle swarm algorithms [18][19], the HS method is a random search technique.It does not require any prior domain knowledge, such as the gradient information of the objective functions.Unfortunately, empirical study has shown that the classical HS method sometimes suffers from a slow search speed, and it is not suitable for handling the multi-modal problems [17].
The steps in the procedure of classical harmony search algorithm are as follows: Step 1. Initialize the problem and algorithm parameters.
Step 2. Initialize the harmony memory.
Step 3. Improvise a new harmony.
Step 4. Update the harmony memory.
Step 5. Check the stopping criterion.These steps are described in the next five subsections.

Initialize the problem and algorithm parameters
In Step 1, the optimization problem is specified as follows: Minimize ( ) where ( ) f x is an objective function; x is the set of each decision variable i x ; N is the number of decision variables, i X is the set of the possible range of values for each decision variable, . The HS algorithm parameters are also specified in this step.
These are the harmony memory size (HMS), or the number of solution vectors in the harmony memory; harmony memory considering rate (HMCR); pitch adjusting rate (PAR); and the number of improvisations (Tmax), or stopping criterion.The harmony memory (HM) is a memory location where all the solution vectors (sets of decision variables) are stored.This HM is similar to the genetic pool in the GA.Here, HMCR and PAR are parameters that are used to improve the solution vector.Both are defined in step 3.

Initialize the harmony memory
In Step 2, the HM matrix is filled with as many randomly generated solution vectors as the HMS

Improvise a new harmony
( , , , ) , is generated based on three rules: memory consideration, pitch adjustment and random selection.Generating a new harmony is called 'improvisation'.In the memory consideration, the value of the first decision variable ' 1 x for the new vector is chosen from any of the values in the specified HM.The HMCR, which varies between 0 and 1, is the rate of choosing one value from the historical values stored in the HM, while (1-HMCR) is the rate of randomly selecting one value from the possible range of values.
( , , , ), , Where rand is a random number between 0 and 1.For example, a HMCR of 0.85 indicates that the HS algorithm will choose the decision variable value from historically stored values in the HM with an 85% probability or from the entire possible range with a (100-85)% probability.Every component obtained by the memory consideration is examined to determine whether it Where bw is an arbitrary distance bandwidth; rand is a random number between 0 and 1.
In Step 3, HM consideration, pitch adjustment or random selection is applied to each variable of the new harmony vector in turn.

Update harmony memory
If the new harmony vector, ( , , , ) is better than the worst harmony in the HM, judged in terms of the objective function value, the new harmony is included in the HM and the existing worst harmony is excluded from the HM.

Check stopping criterion
If the stopping criterion (maximum number of improvisations) is satisfied, computation is terminated.Otherwise, Steps 3 and 4 are repeated.

Improved HS Algorithm With Chaos
Experiments with the standard HS algorithm over the benchmark problems show that the algorithm suffers from the problem of premature and/or false convergence, slow convergence especially over multimodal fitness landscape.
To enrich the searching behavior and to avoid being trapped into local optimum, chaotic dynamics is incorporated into the above HS algorithm.The well-known logistic equation (Liu et al., 2005), which exhibits the sensitive dependence on initial conditions, is employed for constructing hybrid HS.The logistic equation is defined as follows.
That is, it exhibits the sensitive dependence on initial conditions, which is the basic characteristic of chaos.A minute difference in the initial value of the chaotic variable would result in a considerable difference in its long time behavior.The track of chaotic variable can travel ergodically over the whole search space.In general, the above chaotic variable has special characters, i.e. ergodicity, pseudo-randomness and irregularity.The process of the chaotic local search could be defined through the following equation: ( (1 ), 1,2, , .
where i cx is the ith chaotic variable, and k denotes the iteration number.Obviously, ( ) k i cx is distributed in the range (0,1) under the conditions that (0) (0,1) \ {0.25, 0.5, 0.75} i cx  .Based on the proposed HS algorithm with the chaotic local search, a hybrid HS with chaos strategy named HSCH is proposed, in which HS is applied to perform global exploration and chaotic local search is employed to perform locally oriented search (exploitation) for the solutions resulted by the HS [23].The procedure of HSCH is described in Figure 1.
The harmony of worst fitness in the HM needs to be improved.We add chaotic local search to pitch adjusting of HS algorithm.Besides, to maintain population diversity, several new harmony are generated by chaos with certain probability (RGR) and incorporated in the new HM.Thus, the resulting HM members are expected to have better fitness than that of the original ones.This strategy can also overcome the premature shortcoming of the regular HS method.Figure 2 shows the computation procedure of the HSCH Algorithm.

Computational Results and Analysis
In this section we perform some numerical tests in order to illustrate the implementation and efficiency of the proposed method.All the experiments were performed on Windows XP system running on a Hp540 laptop with Intel(R) Core(TM) 2×1.8GHz and 2GB RAM, and the codes were written in MatlabR2010b.

AVE Problems AVE 1.
Let A be a matrix whose diagonal elements are 500 and the nondiagonal elements are chosen randomly from the interval [1,2] where I is the identity matrix of order n and e is 1  n vector whose elements are all equal to unity.Since singular values matrix A are all greater than 1, this AVE is uniquely solvable by Lemma 2.2, and the unique solution is where I is the identity matrix order n and e is 1  n vector whose elements are all equal to unity.Since singular values matrix A are all greater than 1, this AVE is uniquely solvable by Lemma 2.2, and the unique solution is Since singular values matrix A are all greater than 1, this AVE is uniquely solvable by Lemma 2.2, and the unique solution is (1, 1, ,1) T x   .

Results
To judge the accuracy of different algorithms, 30 independent runs of each of the three algorithms were carried out and the best, the mean, the worst fitness values, and the standard deviation (Std) were recorded.Table 1 and Table 2 compares the algorithms on the quality of the optimum solution for every AVE problem of n=50 and n=100.Figure 4 and Figure 5 show the convergence and its boxplot of the best fitness in the population for the different algorithms (HSCH, HS, HSDE) for every AVE problem of n=50 and n=100.The values plotted for every generation are averaged over 30 independent runs.The boxplot is the best fitness in the final population for the different algorithms (HSCH, HS, HSDE) for corresponding AVE problem.

Analysis
We can see that in all instances the HSCH algorithm performs extremely well, and finally converges to the unique solution of the AVE.The behavior of the two former algorithms (HS and HSDE) is similar for all given AVE problems.From the computation results, the HSDE algorithm is clearly the worst algorithm for all given AVE problems, while the HSCH algorithm is the best.

Conclusion
We have proposed HSCH algorithm for solving absolute value equation Ax -|x| = b under the condition that the singular values of A exceed 1.The HM members in HSCH algorithm are fine-tuned by the chaos operator to improve their affinities so that enhanced optimization performances can be achieved.This ensures that the explorative power of HSCH is on average greater than that of HS and HSDE, which in turn results into better accuracy of the HSCH algorithm.

Lemma 2 . 2 (
Mangasarian, AVE with unique solution).(i) The AVE (1) is uniquely solvable for any n b R  if singular values of A exceed 1.

( 1 ,Figure 3 . 3 .
Figure 3. Generating data (A, b) of AVE1 and AVE2 by the Matlab scripts Several simulation examples have been used to verify the effectiveness of the proposed methods.Compared with the HS and HSDE, better optimization results are obtained  ISSN: 1693-6930 TELKOMNIKA Vol.11, No. 4, December 2013 : 835 -844 842 by using HSCH approaches.Future works will also focus on studying the applications of HSCH on engineering optimization problems.

Figure 4 . 843 Figure 5 .
Figure 4.The convergence and boxplot of the best fitness for given AVE problems of n=50 This operation uses the PAR parameter, which is the rate of pitch adjustment as follows:Pitch adjusting decision for  ISSN: 1693-6930 TELKOMNIKA Vol.11, No. 4, December 2013 : 835 -844 838 should be pitch-adjusted.
841were set the same parameters: harmony memory size HMS=15, harmony memory consideration rate HMCR= 0.6, and the number of improvisations Tmax=10000.In HSCH, RGR is chosen to be 0.2.In classical HS, we set pitch adjusting rate PAR= 0.35.Let the initial point ISSN: 1693-6930  Improved Harmony Search Algorithm with Chaos for Absolute Value Equation (Longquan Yong)

Table 1 .
The Statistical Results for 30 Runs on Three given AVE Problems of n=50

Table 2 .
The Statistical Results for 30 Runs on Three given AVE Problems of n=100