 Research
 Open access
 Published:
The qG method
A qversion of the Steepest Descent method for global optimization
SpringerPlus volume 4, Article number: 647 (2015)
Abstract
In this work, the qGradient (qG) method, a qversion of the Steepest Descent method, is presented. The main idea behind the qG method is the use of the negative of the qgradient vector of the objective function as the search direction. The qgradient vector, or simply the qgradient, is a generalization of the classical gradient vector based on the concept of Jackson’s derivative from the qcalculus. Its use provides the algorithm an effective mechanism for escaping from local minima. The qG method reduces to the Steepest Descent method when the parameter q tends to 1. The algorithm has three free parameters and it is implemented so that the search process gradually shifts from global exploration in the beginning to local exploitation in the end. We evaluated the qG method on 34 test functions, and compared its performance with 34 optimization algorithms, including derivativefree algorithms and the Steepest Descent method. Our results show that the qG method is competitive and has a great potential for solving multimodal optimization problems.
Background
The history of qcalculus dates back to the beginning of the previous century when, based on the pioneering works of Euler and Heine, the English reverend Frank Hilton Jackson developed qcalculus in a systematic way (Chaundy 1962; Ernst 2000; Kac and Cheung 2002). His work gave rise to generalizations of special numbers, series, functions and, more importantly, to the concepts of the qderivative (Jackson 1908), or Jackson’s derivative, and the qintegral (Jackson 1910). Recently, based on Jackson’s derivative, a qversion of the classical Steepest Descent method, called the qGradient (qG) method, has been proposed for solving unconstrained continuous global optimization problems (Soterroni et al. 2010, 2012). The main idea behind this new method is the use of the negative of the qgradient of the objective function as the search direction. The qgradient is calculated based on qderivatives, or Jackson’s derivatives, and requires a dilation parameter q that controls the balance between global and local search.
The gradient of onevariable function f(x) is simply the derivative. Geometrically, it is the slope of the tangent line at a given point x; see Fig. 1. Similarly, the qgradient of f is the qderivative that has also a straightforward geometric interpretation as the slope of the secant line passing through the points [x, f(x)] and [qx, f(qx)]. It is immediately evident that the sign of the qderivative can be either positive or negative, depending on the value of the parameter q. For \(q_1\) and \(q_2\) (Fig. 1), the sign of the qderivative is positive and the qG method will move to the left as the Steepest Descent method would do. However, for \(q_3\) the sign of the qderivative is negative which potentially allows the qG method to move to the right direction, towards the global minimum of f.
Figure 2 shows the contour lines of a quadratic function \(f(\mathbf {x})\) with \(\mathbf {x} \in \mathbb {R}^2\), shifted to \(\mathbf {x}=(10,10)\), and the negative of both classical and qgradient vectors. The negative of the classical gradient vector of f, the steepest descent direction, is represented in Fig. 2 by the bold line segment of number 1. All the other line segments illustrate the negative of the qgradient vector obtained for different values of the parameter \(\mathbf {q}\in \mathbb {R}^2\), \(\mathbf {q}=(q_1,q_2)\). The line segments from 2 to 7 use positive values of \(q_1\) and \(q_2\) taken symmetrically around 1 [\(\mathbf {q}=(q_1,q_2\))], and the line segments from 2’ to 7’ use the same values of \(q_1\) and \(q_2\), but with inverted positions [\(\mathbf {q}=(q_2,q_1)\)]. The line segments from 8 to 10 and 8′ to 10′ have a similar description, but use negative values of \(q_1\) and \(q_2\). Note that as the parameter \(\mathbf {q}\) approaches \(\mathbf {1}\), the negative of the qgradient tends to the steepest descent direction. As can be seen, depending on the value of the parameter q, the qgradient can point to any direction and not only to descent directions.
Figure 3 illustrates the points sampled by the qG method, with two different strategies to generate the parameter q, over a multimodal function \(f(\mathbf {x})\), \(\mathbf {x} \in \mathbb {R}^2\). The initial point of the search is \(\mathbf {x}^0=(11,11)\) and the strategy adopted to define the step length is the same in both cases. Note that the function f has a local minimum at \(\mathbf {x}=(10,10)\) and a global minimum at \(\mathbf {x}=(13,13)\). In Fig. 3a, the parameter \(q_i^k\) is fixed and close to 1 along all the iterations k (\(q_i^k \approx 1\), \(\forall i, \quad i=1,2\)) and, consequently, the qG method performs a local search converging to the nearest local minimum to the starting point. In Fig. 3b, the parameters \(q_i^k\) are different from 1 in the beginning and tend to 1 along the iterative procedure (\(q_i^k \rightarrow 1\), \(\forall i, \quad i=1,2\)), with the qG method performing a global search, with exploration of the search space and exploitation of the global minimum vicinity. As can be seen in Fig. 3, the possibility of having different values of the parameter \(q_i^k\), and not only values close to 1, enables the method to move against the local descent direction, escape from the local minimum and sample other regions of the search space. In other words, when the values of \(q_i^k\) are sufficiently different from 1, the qgradient vector can point to any region of the search space. This potentially allows the qG method to move towards the global minimum.
These simple examples show that the use of the qgradient, based on Jackson’s derivative, offers a new mechanism for escaping from local minima. The algorithm for the qG method is complemented with strategies to generate the parameter q and to compute the step length in a way that the search process gradually shifts from global in the beginning to almost local search in the end. As here proposed, the qG algorithm has only two free parameters to be adjusted: the initial standard deviation (\(\sigma ^0\)) and the reduction factor (\(\beta\)). Although a bad choice may lead to some deterioration in its performance, the qG method has shown to be sufficiently robust for still being capable of reaching the global minimum.
We evaluated qG’s performance of against 34 optimization methods and on 34 test problems. First, we considered ten 2D optimization problems, eight unimodal and two multimodal, as defined in (Luksan and Vlcek 2000), and compared the qG method with 22 derivativefree algorithms described in (Rios and Sahinidis 2013). Second, we evaluated our approach on twelve 10D and twelve 30D test problems, ten unimodal and fourteen multimodal. These test problems have been proposed as a benchmark for the CEC′2005 Special Session on RealParameter Optimization of the IEEE Congress on Evolutionary Computation 2005 (Suganthan et al. 2005). For this second set of problems, we compared the qG with 11 Evolutionary Algorithms (EAs) participants of the competition and with the steepest descent (SD) method.
The rest of the paper is organized as follows. The " qgradient vector" section introduces the qgradient vector. "The qG method" section presents an algorithm for the qG method. "Computational experiments" section shows the numerical results, and "Conclusions" section presents our main conclusions.
The qgradient vector
Let f(x) be a differentiable onevariable function. The classical derivative of f is given by
and it is related to infinitesimal summation of the independent variable x. The qderivative is related to dilations of the independent variable x that is multiplied by the parameter q as follows (Jackson 1908)
where \(x\ne 0\) and \(q\ne 1\). The parameter q can be any number different from 1. Therefore, the qderivative can be written as (Koekoev and Koekoev 1993)
Similarly, let \(f(\mathbf {x})\) be a differentiable function of n real variables, then the firstorder partial qderivative of f with respect to the variable \(x_i\) can be given by (Soterroni et al. 2010)
Thus, the qgradient is the vector of the n firstorder partial qderivatives of f defined as (Soterroni et al. 2010)
where the parameter q is a vector \(\mathbf {q} = (q_1,\ldots ,q_i,\ldots ,q_n) \in \mathbb {R}^n\). The classical gradient is recovered in the limit of \(q_i \rightarrow 1\), for all \(i=1,...,n\).
The qG method
Let a general nonlinear unconstrained optimization problem be defined as
where \(\mathbf {x} \in \mathbb {R}^n\) is the vector of the independent variables and \(f:\mathbb {R}^n \rightarrow \mathbb {R}\) is the objective function. A common optimization strategy is to consider an iterative procedure that, starting from an initial point \(\mathbf {x}^0\), generates a sequence \(\{\mathbf {x}^k \}\) given by
where k is the iteration number, \(\mathbf {d}^{k}\) is the search direction and \(\alpha ^{k}\) is the step length or the distance moved along \(\mathbf {d}^{k}\) in the iteration k.
Gradientbased optimization methods can be characterized by the different strategies used to move through the search space. The Steepest Descent method, for example, sets \(\mathbf {d}^{k} = \nabla f(\mathbf {x}^{k})\) as the search direction and the step length \(\alpha ^{k}\) is usually determined by a linesearch technique that minimizes the objective function along the direction \(\mathbf {d}^{k}\). In the qG method, the search direction is the negative of the qgradient of the objective function \( \nabla _{q} f(\mathbf {x})\).
Therefore, the optimization procedure defined by (7) becomes
Key to the performance of the qG method is the correct specification of the parameter q. Considering a function of n variables \(f(\mathbf {x})\), a set of n different parameters \(q_i \in \mathbb {R}\{1\}\) (\(i=1,\ldots ,n\)) are needed to compute the qgradient vector of f. The strategy adopted here is to draw the values of \(q_ix_i\), \(\forall i, \ i=1,\ldots ,n\), from a Gaussian probability density function (pdf), with a standard deviation that decreases as the iterative search proceeds and the mean equal to the current point of the search \(x_i\).
Starting from \(\sigma ^0\), the standard deviation of the pdf is decreased by the following “cooling” schedule, \(\sigma ^{k+1} = \beta \cdot \sigma ^{k}\), where \(0<\beta <1\) is the reduction factor (see Fig. 4d). As \(\sigma ^{k}\) approaches zero, the values of \(q_i^k\) tend to 1, the algorithm reduces to the Steepest Descent method, and the search process becomes essentially local. In this sense, the role of the standard deviation here is reminiscent of the one played by the temperature in a Simulated Annealing (SA) algorithm, that is, to make the algorithm sweeps from a global random sampling in the beginning to a local deterministic search in the end. As in a SA algorithm, the performance of the minimization algorithm depends crucially on the choice of parameters \(\sigma ^0\) and \(\beta\). A too rapid decrease of \(\sigma ^{k}\), for example, may cause the algorithm to be trapped in a local minimum.
Figure 4a–c show 500 line segments that represent the possible search directions of the qG method for a 2D quadratic function at \(\mathbf {x}^0=(11,11)\). The parameters q were generated by a Gaussian distribution with standard deviation \(\sigma ^k\) that decreases along the iterative procedure as \(\sigma ^{k+1}=\beta \cdot \sigma ^k\), with a initial standard deviation \(\sigma ^0\) and the reduction factor \(\beta\). In the beginning of the iterative procedure (Fig. 4a), the search directions can point to anywhere but, as the \(\sigma ^k\) tends to 0, the parameter \(q_i^k\) tends to 1 and, consequently, the search directions tend to the steepest descent direction (Fig. 4c).
The algorithm of the qG method is completed with a strategy for calculating the step length \(\alpha ^k\) based on standard parabolic interpolation. Given the current point of the search \(\mathbf {x}^k\) and the parameter \(\mathbf {q}^k\), we calculate \(\gamma = \Vert \mathbf {q}^k \mathbf {x}^k  \mathbf {x}^k\Vert\) to define the triplet of points \((\mathbf {a},\mathbf {b},\mathbf {c})=(\mathbf {x}^k  \gamma \mathbf {d}^k,\mathbf {x}^k, \mathbf {x}^k + \gamma \mathbf {d}^k )\), where \(\mathbf {d}^k\) is the search direction in the iteration k. The step length corresponds to the distance between the current value and the minimum value of the parabola passing through \(\mathbf {a}\), \(\mathbf {b}\) and \(\mathbf {c}\). Other strategies for computing the step length are, naturally, possible. Nevertheless, the use of a linesearch technique is not recommended since the search directions of the qG method are not always descent directions. Descent directions are required for the success of this kind of onedimensional minimization (Nocedal and Wright 2006).
Summarizing, the algorithm of the qG method for unconstrained continuous global optimization problems is described as follows.
Algorithm for the qG method
Given \(\mathbf {x}^0\) (initial point), \(\sigma ^0>0\) and \(\beta \in (0,1)\):

1.
Set \(k= 0\)

2.
Set \(\mathbf {x}_{best} = \mathbf {x}^k\)

3.
While the stopping criteria are not reached, do:

a.
Generate \(\mathbf {q}^k\mathbf {x}^k\) by a Gaussian distribution with \(\sigma ^k\) and \(\mu ^k=\mathbf {x}^k\)

b.
Calculate the qgradient vector \(\nabla _q f(\mathbf {x}^k)\)

c.
Set \(\mathbf {d}^{k} =  \nabla _q f(\mathbf {x}^k) / \Vert \nabla _q f(\mathbf {x}^k) \Vert\)

d.
Calculate \(\gamma = \Vert \mathbf {q}^k \mathbf {x}^k  \mathbf {x}^k\Vert\)

e.
Calculate the triplet \((\mathbf {a},\mathbf {b},\mathbf {c})=(\mathbf {x}^k  \gamma \mathbf {d}^k,\mathbf {x}^k, \mathbf {x}^k + \gamma \mathbf {d}^k )\)

f.
Compute \(\alpha ^k\) by parabolic interpotation, and \(\mathbf {x}^{k+1}= \mathbf {x}^{k} + \alpha ^{k} \mathbf {d}^{k}\)

g.
If \(f(\mathbf {x}^{k+1}) < f(\mathbf {x}_{best})\) set \(\mathbf {x}_{best} = \mathbf {x}^{k+1}\)

h.
Set \(\sigma ^{k+1} = \beta \cdot \sigma ^{k}\)

i.
Set \(k = k + 1\)

a.

4.
Return \(\mathbf {x}_{best}\) and \(f(\mathbf {x}_{best})\).
The qG method stops when the appropriate stopping criterion is attained. In realworld applications (i.e., in problems for which the global minimum is not known), it can be the maximum number of function evaluations, or the value of the local gradient \(\nabla f(\mathbf {x}^k) < \epsilon\) (\(\epsilon >0\)), since the qG method converges to the Steepest Descent method. The algorithm returns the \(\mathbf {x}_{best}\) as the minimum value of the objective function f obtained during the iterative procedure, i.e., \(f(\mathbf {x}_{best}) \le f(\mathbf {x}^k)\), \(\forall k\).
Computational experiments
The qG method was tested on two set of problems followed by a systematic comparison with derivativefree algorithms. First, we applied our approach on ten 2D test problems defined in (Luksan and Vlcek 2000), eight unimodal and two multimodal, and compared it with 22 derivativefree algorithms described in (Rios and Sahinidis 2013). Second, the qG method was evaluated on twelve 10D and twelve 30D test problems, ten unimodal and fourteen multimodal, and it was compared with 11 Evolutionary Algorithms (EAs) participants of the CEC′2005 Special Section on RealParameter Optimization of the 2005 IEEE Congress on Evolutionary Computation, and the Steepest Descent method. The qG method and the Steepest Descent method were performed on a iMac 2.7GHz with Intel Core i5 processor and 8GB RAM running Intel Fortran Composer XE for Mac OS* X.
The qG method has two free parameters, the initial standard deviation \(\sigma ^0\) and the reduction factor \(\beta\). The optimal setting for these parameters is typically problem dependent. The value of \(\sigma ^0\) should not be too small so that the algorithm does not behave like its classical version too early, and not too large so that the iterates do not escape the search space frequently. Extensive numerical tests have shown that \(\sigma ^0 = \kappa L\), where \(\kappa = \sqrt{D/2}\), D is the dimension of the problem and L is the largest distance within the search space given by \(L=\sqrt{\sum _{i=1}^{n} (\mathbf {x}_{{max}_{i}}  \mathbf {x}_{{min}_{i}})}\), provides a simple heuristic for setting this parameter. For example, we set \(\sigma ^0 =L\) for all the 2D problems. This strategy may not provide the best parameter setting for every test function, but is good enough for a wide range of problems.
The value of \(\beta\), which controls the speed with which the algorithm shifts from global to local search, also depends on D. As a rule of thumb, \(\log (1\beta ) \sim \kappa\) gives good estimates for \(\beta\). For example, for D ranging from 1 to 8, \(0.9 \le \beta \le 0.99\) is a good choice; for \(8 \le D \le 18\), we choose \(0.99 \le \beta \le 0.999\); for \(18 \le D \le 32\), \(0.999 \le \beta \le 0.9999\); and so on. Naturally, the choice of \(\beta\) should be balanced with the maximum number of function evaluations of one is willing or limited to perform.
First set of problems
The first ten 2D test problems, eight unimodal and two multimodal, are described in Table 1. The evaluation of the objective function at the global optimum \(f(\mathbf {x}^{*})\) (last column of Table 1) is used to define a successful run. A run is considered successful if the objective function value at \(\mathbf {x}\) is within \(1\%\) or 0.01 of the global optimum (Rios and Sahinidis 2013). For each problem, 10 independent runs were performed from random starting points obtained from a uniform distribution within the search space, as defined in (Rios and Sahinidis 2013). The same starting points were used for all methods, including our approach. The stopping criterion used by all algorithms is \(N_{max}=2500\) evaluations of the objective function per run.
The resolution of all 22 derivativefree methods and the qG method over this set of ten problems, each one solved for 10 independent runs, results in a total number of 100 optimization instances per algorithm. In order to perform a qualitative comparison of the different methods over this set of problems, we calculate for each algorithm the fraction of solved problems given by the ratio between the number of successful runs and the total number of instances. This ratio is computed every 50 evaluations of the objective function. Figures 5, 6 and 7 show the fraction of multimodal, unimodal and all problems, respectively, solved by each method to reach the optimality tolerance along the iterative procedure.
The qG method solved 100 % of the multimodal problems (see Fig. 5), 79 % of the unimodal problems (see Fig. 6) and, in total, our approach solved 83 % of the problems arriving in a 7th position among the 23 methods. The comparison was performed under the same initial conditions and stopping criterion, and the qG method used the same set of free parameters for all problems, namely \(\sigma ^0 = L\) and \(\beta = 0.95\).
Second set of problems
The performance of the qG method was also evaluated on twelve 10D and twelve 30D test problems, ten unimodal and fourteen multimodal. Their characteristics are summarized in Table 2. These functions are a subset of the problems proposed at CEC′2005 (Suganthan et al. 2005). The qG method is compared with the top eleven Evolutionary Algorithms (EAs) participants of the competition: BLXGL50 (GarciaMarinez and Lozano 2005), BLXMA (Molina et al. 2005), CoEVO (Posik 2005), DE (Ronkkonen et al. 2005), EDA (Yuan and Gallagher 2005), DMSLPSO (Liang and Suganthan 2005), LSaDE (Qin and Suganthan 2005), GCMAES (Auger and Hansen 2005a), LCMAES (Auger and Hansen 2005b), KPCX (Sinha et al. 2005) and SPCPNX (Ballester et al. 2005). In addition, we applied the Steepest Descent (SD) method to the same test problems in order to compare the qG method with its classical version.
The test problems proposed at CEC′2005 are based on classical benchmark functions such as Sphere, Rosenbrock’s, Rastrigin’s, Ackley’s and Griewank’s function. They contain difficulties such as huge number of local minima, shifted global optimum, rotated domain, noise, global optimum outside the initialization range or within a very narrow basin, and a combination of different function properties (Suganthan et al. 2005). The function \(f_{15}\), for example, is a composition of Rastrigin’s, Weierstrass, Griewank’s, Ackley’s and Sphere functions. The selected subset of the CEC′2005 test problems comprises those for which at least one of the thirteen algorithms (eleven EAs, the qG and SD methods) was capable to achieve a fixed accuracy or a target function value defined in (Suganthan et al. 2005).
To ensure a fair comparison, we applied on the qG and SD methods the same evaluation criteria defined in (Suganthan et al. 2005) and used by the EAs. For each function and dimension, the algorithms performed 25 independent runs from different starting points generated with a uniform random distribution within the search space^{Footnote 1} (see column \([\mathbf {x}_{min},\mathbf {x}_{max}]^D\) of Table 2). The stopping criteria are either the termination error value equal to \(10^{8}\) or less {i.e., \([f(\mathbf {x}_{best})f(\mathbf {x}^{*})] < 10^{8}\), where \(\mathbf {x}^{*}\) is the global optimum} or the maximum number of function evaluations (\(Max\_FEs\)) equal to \(10{,}000 \times D\). More details of the functions and evaluation criteria can be found in (Suganthan et al. 2005). The resolution of these twelve 10D and twelve 30D problems, each one solved for 25 independent runs, results in a total number of 600 optimization instances per algorithm. Here we are also using fixed free parameters \(\sigma ^0 = \sqrt{5}L\) with \(\beta = 0.995\) for all 10D problems, and \(\sigma ^0 =\sqrt{15}L\) with \(\beta = 0.9995\) for all 30D problems. The SD method uses golden section technique to generate the step length. For the multimodal functions, whenever the SD method terminates before reaching the termination error and the number of function evaluations is less than \(Max\_FEs\), it is restarted from a randomly generated point inside the search space.
To evaluate the performance of the methods on this second set of problems, we computed a “success rate” (SR) and a “success performance” (SP) for each algorithm and function as (Suganthan et al. 2005)
SR represents the fraction of runs that were successful. A successful run for this set of problem is defined as a run in which the algorithm achieves a given accuracy level^{Footnote 2} with less or equal \(Max\_FEs\) (Suganthan et al. 2005). SP is the number of function evaluations needed for an algorithm to achieve the minimum, with a given accuracy, in a successful run. Table 3 presents SR and SP values obtained by the qG and SD methods, for all the twelve 10D and twelve 30D test functions.
The performance comparison of the qG method with its classical version (SD method) and the top eleven EAs of CEC′2005 was made for two groups: the five unimodal and the eight multimodal test functions. For each group and dimensions, 10D and 30D, we calculated the average success rate and the average success performance of each algorithm. The average success rate of an algorithm is the arithmetic mean of the SP values in a group of functions. The average success performance is the arithmetic mean of the SP values for the functions with at least one successful run. Tables 4 and 5 show the resulting ranking of the algorithms for the unimodal and multimodal functions, respectively, and the dimensions 10D and 30D. The algorithms are ranked by the following criteria:

1.
Highest value of the average success rate (column “Average SR %”).

2.
Number of solved functions in each group (column “SF”). A function is considered solved by an algorithm if at least one of the runs is a successful run or if the SR is different from 0.

3.
Lowest value of the average success performance (column “Average SP”).
For the unimodal problems, the qG method does not perform very well arriving in eleventh and tenth positions for dimensions 10D and 30D, respectively. The average success rates for the unimodal problems are 59 % for 10D and 20 % for 30D. Note that the increase of the dimension affected the performance of all algorithms, in terms of either SR or the number of solved functions. Overall, the performance of the qG method is not very different from its classical version, the SD method, which arrives in the thirteenth and ninth positions, for dimensions 10D and 30D, respectively.
This picture changes for the multimodal problems, where the qG method performed well, arriving in fifth and second positions for 10D and 30D, respectively. The average success rates of the qG method are 43 and 41 % for 10D and 30D, respectively. As expected, the SD method has a poor performance over the multimodal problems arriving in twelfth and eighth positions for dimensions 10D and 30D, respectively. Again, the increase of the dimension affected the performance of all algorithms.
Conclusions
In this paper we presented the qG method, a generalization of the Steepest Descent method based on the use of the qgradient vector to compute the search direction. This strategy provides the algorithm an effective mechanism for escaping from local minima. As implemented here, the search process performed by the qG method gradually shifts from global search in the beginning to local search in the end. Our computational results have shown that the qG method is competitive and promising. For the multimodal functions in the two set of problems, it performed well compared to the other derivativefree algorithms, some considered to be among the stateoftheart in the evolutionary computation and numerical optimization communities.
Although our preliminary results show that the method is effective, further research is necessary. Currently, a novel version of the qG method, which is able to guarantee the convergence of the algorithm to the global minimum in a probabilistic sense, is under development. This version is based on the generalized adaptive random search (GARS) framework for deterministic functions (Regis 2010). In addition, gains in the performance of the qG method are expected with the implementation of several improvements, such as inclusion of side, linear and nonlinear restrictions, development of better step selection strategies and others.
Notes
For the Shifted rotated Griewank’s function (\(f_7\)) the domain used to generate the initial points is \([0,600]^D\), where D is the dimension of the problem.
The accuracy level is defined for each function in (Suganthan et al. 2005).
References
Auger A, Hansen N (2005) A restart CMA evolution strategy with increasing population size. In: Proceedings of 2005 IEEE Congress on Evolutionary Computation, vol 2. Edinburgh, UK, pp 1769–1776
Auger A, Hansen N (2005) Performance evaluation of an advanced local search evolutionary algorithm. In: Proceedings of 2005 IEEE Congress on Evolutionary Computation, vol 2. Edinburgh, UK, pp 1777–1784
Ballester PJ, Stephenson J, Carter JN, Gallagher K (2005) Realparameter optimization performance study on the CEC2005 benchmark with SPCPNX. In: Proceedings of 2005 IEEE Congress on Evolutionary Computing, vol 1. Edinburgh, UK, pp 498–505
Ernst T (2000) The history of \(q\)calculus and a newmethod. U.U.D.M. report 2000:16, Department of Mathematics, Uppsala University, Sweden
Chaundy TW (1962) Frank Hilton Jackson (obituary). J Lond Math Soc 37:126–128
GarciaMartinez C, Lozano M (2005) Hybrid realcoded genetic algorithms with female and male differentiation. In: Proceedings of 2005 IEEE Congress Evolutionary Computing, vol 1. Edinburgh, UK, pp 896–903
Jackson FH (1908) On \(q\)functions and a certain difference operator. Trans R Soc Edinb 46:253–281
Jackson FH (1910) On \(q\)definite integrals. Quart J Pure Appl Math 41:193–203
Kac V, Cheung P (2002) Quantum calculus (Universitext). Springer, New York
Koekoev J, Koekoev R (1993) A note on the qderivative operator. J Math Anal Appl 176:627–634
Liang JJ, Suganthan PN (2005) Dynamic multiswarm particle swarm optimizer with local search. In: Proceedings of 2005 IEEE Congress Evolutionary Computing, vol 1. Edinburgh, UK, pp 522–528
Luksan L, Vlcek J (2000) Test problems for nonsmooth unconstrained and linearly constrained optimization. Technical report, Institute of Computer Science, Academy of Sciences of the Czech Republic. http://www3.cs.cas.cz/ics/reports/v79800.ps. Accessed 01 Sept 2015
Molina D, Herrera F, Lozano M (2005) Adaptative local search parameters for realcoded memetic algorithms. In: Proceedings of 2005 IEEE Congress Evolutionary Computing, vol 1. Edinburgh, UK, pp 888–895
Nocedal J, Wright SJ (2006) Numerical optimization, 2nd edn. Springer Science Business Media, New York,
Posik P (2005) Realparameter optimization using the mutation step coevolution. In: Proceedings of 2005 IEEE Congress Evolutionary Computing, vol 1. Edinburgh, UK, pp 872–879
Qin AK, Suganthan PN (2005) Selfadaptive differential evolution algorithm for numerical optimization. In: Proceedings of 2005 IEEE Congress Evolutionary Computing, vol 2. Edinburgh, UK, pp 1785–1791
Regis RG (2010) Convergence guarantees for generalized adaptive stochastic search methods for continuous global optimization. Eur J Oper Res 207(3):1187–1202
Rios LM, Sahinidis NV (2013) Derivativefree optimization: a review of algorithms and comparison of software implementations. J Glob Optim 56(3):1247–1293
Ronkkonen J, Kukkonen S, Price KV (2005) Realparameter optimization with differential evolution. In: Proceedings of 2005 IEEE Congress Evolutionary Computing, vol 1. Edinburgh, UK, pp 506–513
Sinha A, Tiwari S, Deb K (2005) A populationbased, steadystate procedure for realparameter optimization. In: Proceedings of 2005 IEEE Congress Evolutionary Computing, vol 1. Edinburgh, UK, pp 514–521
Soterroni AC, Galski RL, Ramos FM (2011) The \(q\)gradient vector for unconstrained continuous optimization problems. In: Bo H et al (ed) Operations research proceedings 2010. SpringerVerlag Berlin Heidelberg, pp 365–370. doi:10.1007/9783642200090_58
Soterroni AC, Galski RL, Ramos FM (2012) The \(q\)gradient method for global optimization. arXiv:1209.2084 [math.OC]
Suganthan P, Hansen N, Liang JJ, Deb K, Chen YP, Auger A, Tiwari S (2005) Problem definitions and evaluation criteria for the CEC 2005 special session on realparameter optimization. Technical report, Nanyang Technological University and Kanpur Genetic Algorithms Laboratory. http://web.mysites.ntu.edu.sg/epnsugan/PublicSite/Shared%20Documents/CEC2005/TechReportMay3005. Accessed 01 Sept 2015
Yuan B, Gallagher M (2005) Experimental results for the special session on realparameter optimization at CEC 2005: a simple, continuous EDA. In: Proceedings of 2005 IEEE Congress on Evolutionary Computing, vol 2. Edinburgh, UK, pp 1792–1799
Authors’ contributions
ACS, RLG, MCS and FMR participated in the design of the study and performed the statistical analysis. They also conceived of the study, and participated in its design and coordination and helped to draft the manuscript. All authors read and approved the final manuscript.
Acknowledgements
The authors gratefully acknowledge the support provided by the National Counsel of Technological and Scientific Development (CNPq) and Coordination for the Improvement of Higher Education Personnel (CAPES), Brazil.
Competing interests
The authors declare that they have no competing interests.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
About this article
Cite this article
Soterroni, A.C., Galski, R.L., Scarabello, M.C. et al. The qG method . SpringerPlus 4, 647 (2015). https://doi.org/10.1186/s4006401514344
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s4006401514344