CN116306919A - Large-scale multi-objective combination optimization method based on problem recombination and application - Google Patents
Large-scale multi-objective combination optimization method based on problem recombination and application Download PDFInfo
- Publication number
- CN116306919A CN116306919A CN202310233019.XA CN202310233019A CN116306919A CN 116306919 A CN116306919 A CN 116306919A CN 202310233019 A CN202310233019 A CN 202310233019A CN 116306919 A CN116306919 A CN 116306919A
- Authority
- CN
- China
- Prior art keywords
- optimization
- population
- convergence
- objective
- diversity
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000005457 optimization Methods 0.000 title claims abstract description 105
- 238000000034 method Methods 0.000 title claims abstract description 58
- 238000005215 recombination Methods 0.000 title claims abstract description 25
- 230000006798 recombination Effects 0.000 title claims abstract description 24
- 230000008569 process Effects 0.000 claims abstract description 16
- 230000007246 mechanism Effects 0.000 claims abstract description 12
- 238000005516 engineering process Methods 0.000 claims abstract description 3
- 239000013598 vector Substances 0.000 claims description 33
- 239000000243 solution Substances 0.000 claims description 25
- 239000012088 reference solution Substances 0.000 claims description 17
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000004891 communication Methods 0.000 claims description 6
- 238000005265 energy consumption Methods 0.000 claims description 6
- 230000010354 integration Effects 0.000 claims description 6
- 230000035772 mutation Effects 0.000 claims description 5
- 210000000349 chromosome Anatomy 0.000 claims description 3
- 230000007613 environmental effect Effects 0.000 claims description 3
- 108090000623 proteins and genes Proteins 0.000 claims description 3
- 238000009827 uniform distribution Methods 0.000 claims description 2
- 230000000295 complement effect Effects 0.000 abstract description 3
- 238000004458 analytical method Methods 0.000 description 4
- 230000009466 transformation Effects 0.000 description 4
- 238000013459 approach Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000009977 dual effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000002028 premature Effects 0.000 description 1
- 230000008521 reorganization Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Theoretical Computer Science (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Genetics & Genomics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Physiology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a large-scale multi-objective combination optimization method based on problem recombination and application thereof, wherein the method comprises the following steps: randomly initializing a population; based on a decision variable clustering technology, dividing decision variables into convergence variables and diversity variables; adopting directional cross variation facing convergence, and combining a convergence environment selection mechanism to select an optimal individual as a new parent population; when the population encounters selection pressure, problem reconstruction is carried out, and large-scale multi-objective optimization is converted into single-objective optimization; adopting diversity-oriented directional cross variation and combining a diversity environment selection mechanism to select a population as a new parent population; and after the termination condition is met, outputting the parent population as an optimal solution set of the optimization target. Compared with the prior art, the complementary search strategy and the problem reconstruction strategy provided by the invention can solve the problem of large-scale multi-objective optimization, respectively process the problems of convergence and diversity in different optimization stages, and avoid the situation of sinking into local optimum.
Description
Technical Field
The invention relates to the field of multi-objective evolution calculation, in particular to a large-scale multi-objective combined evolution optimization method based on problem recombination.
Background
A variety of multi-objective evolutionary algorithms have emerged over the last two decades, including Pareto-based multi-objective evolutionary algorithms (Multi Objective Evolutionary Algorithms, MOEAs), decomposition-based MOEAs, and index-based MOEAs, among others. While most existing multi-objective evolutionary algorithms exhibit good performance in solving multi-objective problems with a small number of decision variables, their performance drops dramatically when solving multi-objective optimization problems (MOPs) with hundreds or even thousands of decision variables, i.e., large-scale MOPs (LSMOPs). As the number of decision variables increases linearly, the volume (and complexity) of the search space will increase exponentially, resulting in premature convergence of the algorithm to a locally optimal or to an oversized region. In recent years, the academic and industrial industries sequentially put forward to perform optimization research on LSMOPs based on a co-evolution framework, a decision variable analysis and a problem transformation and other multi-objective evolutionary algorithm frameworks, but a plurality of problems to be solved still exist, and the problems are mainly expressed in the following aspects:
LSMOEAs based on co-evolution frameworks take a significant amount of time to analyze decision variables to complete the grouping of decision variables. In addition, when there is an association relationship between sub-problems due to improper grouping, the optimization needs to be repeated in turn, and the performance of the algorithm is also severely degraded. Notably, the separability assumption between decision variables is not always correct. Therefore, the algorithm has limitations and is not suitable for solving large-scale MOPs of each decision variable interaction.
LSMOEAs based on decision variable analysis, although reducing the scale of the problem to some extent through decision variable classification, generate fewer categories (convergence variable, diversity variable and mixed variable), the decomposed sub-problems can still be large-scale problems, and the overall search efficiency of the algorithm is still to be improved.
LSMOEAs based on problem transformation needs to find a problem transformation function to ensure that the information loss after the original problem is transformed into a new problem is as small as possible. However, it is very difficult to find a perfect problem transformation function, and more impossible in particularly complex problems. In addition, since one weight corresponds to a set of decision variables, the search for the decision space is not thorough, and the quality of the obtained final solution is to be improved.
Chinese patent application CN114819040a discloses a dual population co-evolution method based on dual search, which cannot deal with multi-objective optimization problems with decision variable numbers up to 500 or even 1000 or more.
Disclosure of Invention
The invention aims to overcome the defect that the prior art is difficult to balance convergence and diversity in the optimization process, and provides a large-scale multi-objective combination optimization method based on problem recombination.
The aim of the invention can be achieved by the following technical scheme:
as a first aspect of the present invention, there is provided a large-scale multi-objective combinatorial optimization method based on problem reorganization, the method including a convergence optimization stage and a diversity optimization stage, the specific steps including:
randomly initializing a population P;
based on a decision variable clustering technology, dividing decision variables into convergence variables and diversity variables;
convergence optimization stage:
in the evolution process, adopting directional cross variation facing convergence, and combining with a convergence environment selection mechanism, selecting an optimal individual as a parent population P';
when the population encounters selection pressure, problem reconstruction is carried out, large-scale multi-objective optimization is converted into single-objective optimization, and the optimal individuals are reselected as parent populations P';
diversity optimization stage:
in the evolution process, adopting diversity-oriented directional cross variation, and combining a diversity environment selection mechanism to select a population as a new parent population;
and after the termination condition is met, outputting the parent population as an optimal solution set of the optimization target.
Further, the specific evolution steps of the convergence optimization stage are as follows:
using binary competition selection based on convergence, and selecting individuals from the current population to generate N sub-generations Q by using crossover and mutation operators;
calculating the convergence degree of each individual in the current population P and the offspring Q, and selecting the optimal N individuals from the union as a new parent population P';
repeating the above process until the termination condition is satisfied;
performing problem reconstruction, and converting large-scale multi-objective optimization into single-objective optimization through bi-directional weight vector association;
producing a progeny population a by differential evolution;
the convergence policy based environmental selection mechanism selects the optimal N individuals from the union of the offspring population a and the weight vector population Q' as the new parent population P ".
Further, the directional crossing of the convergence performs crossing operation by taking independent convergence variable groups as a unit;
the directional variation of the convergence selects a variable from the convergence variables to vary.
Further, the convergence C of the solution x d The calculation formula is as follows:
where M represents the target number.
Further, the problem recombination strategy specifically includes: through bi-directional weight vector association, the original large-scale multi-objective optimization problem is re-expressed into single-objective optimization with relatively smaller weight variables, and a group of candidate solutions with good convergence and uniform distribution are used for guiding the algorithm to search the direction of the optimal set.
Further, the specific evolution steps of the problem recombination strategy are as follows:
weight vector association, selecting from the current populationr solutions are used as reference solution sets, each reference solution is combined with two direction vectors V l And V u And two weight variables lambda r1 And lambda (lambda) r2 Associating;
constructing a sub-problem according to the direction vector and the weight variable:
Z′(Λ)＝{z 11 (λ 11 )，z 12 (λ 12 )，...，z r1 (λ r1 )z r2 (λ r2 )}
wherein Λ= { λ 11 ，λ 12 ，...，λ r1 ，λ r2 A reconstructed decision space;
target space reconstruction, when the sub-problem is reconstructed, the optimization of the decision vector x in the original decision space is converted into the optimization of the weight vector Λ in the reconstructed decision space, and the new optimization problem is expressed as follows:
maximize G(Λ)＝H(Z′(Λ))
where H may be any performance indicator.
Further, the specific steps of constructing the sub-problem according to the direction vector and the weight variable include:
two direction vectors for each reference solution are calculated:
V l ＝s 1 -o
V u ＝t-s 1
wherein s is 1 ＝{x 1 ，...，x d The } is a reference solution, and t and o are upper and lower boundary points of X respectively;
two weight vectors for each reference solution are calculated:
where λ11 and λ12 are two weight variables, l max ＝||t-o||；
The expression of the constructor problem is:
further, the specific evolution steps of the diversity optimization stage are as follows:
selecting individuals from the current population P' using a binary competition selection based on crowding distances;
utilizing the selected individuals to only cross and mutate the diversity related variables to generate N offspring Q';
selecting half of the population from the union of the current population P ' and the offspring Q ' as a new parent population P ', according to the pareto dominance and crowding distance;
repeating the above process until the termination condition is met, and outputting the parent population P' "as the optimal solution set of the optimization target.
Further, the directional crossover of the diversity performs crossover operation by taking all diversity variables as units;
the directed variation of diversity selects variables among the diversity variables for variation.
As a second aspect of the present invention, there is provided an application of the problem-recombination-based large-scale multi-objective combination optimization method as described in any one of the above, wherein the application scenario of the method includes resource scheduling and integration optimization of a large-scale data center; the specific application steps comprise:
taking indexes including energy consumption, resource loss, data communication flow and task completion time delay as problem optimization targets, taking the residual available computing capacity of a server as constraint conditions, and establishing a multi-target constraint optimization model of a large-scale data center resource scheduling and integration problem;
taking the number of virtual machines to be scheduled as the length of a chromosome, and taking the number of the virtual machines as the gene position, so as to realize the coding of population individuals;
initializing a population P based on a coding mode, performing iterative optimization of the population according to the large-scale multi-objective combination optimization method based on the problem recombination, which is described above, until the termination condition of an algorithm is met, and outputting a Pareto optimal solution set with a plurality of conflict indexes including energy consumption, resource loss, data communication flow and task completion time delay as optimization objectives.
Compared with the prior art, the invention has the following beneficial effects:
1) The large-scale multi-objective optimization method provided by the invention can convert the large-scale multi-objective optimization problem into the optimization of the weight vector based on the problem reconstruction strategy, thereby solving the large-scale problem.
2) Unlike available large scale multi-target evolution algorithm, which maintains the convergence and diversity of population simultaneously during one iterative optimization process, the present invention provides complementary search strategy to treat the convergence and diversity separately in different optimization stages. In the first stage, the population approaches the pareto front rapidly, ignoring the diversity of the population. And after the population converges, the decision variable clustering method is adopted in the second stage to emphasize the diversity of the population, so that the situation of sinking into local optimum is avoided.
3) Different from the existing multi-objective evolutionary algorithm, which adopts methods such as crowding degree and the like in environment selection, the invention provides a convergence degree concept for quantitatively judging the convergence of candidate solutions so as to select a solution with better convergence.
4) The diversity evolution of the invention can adopt other methods of multi-objective optimization algorithms, so that different instantiation algorithms are designed aiming at different types of multi-objective optimization problems, and the expansibility is strong.
Drawings
FIG. 1 is a schematic flow diagram of a large-scale multi-objective combined evolution method based on problem recombination according to the present invention;
FIG. 2 is a schematic diagram of an approximate Pareto front on test function LSMOF2 for an exemplary algorithm of the present invention;
FIG. 3 is a schematic diagram of an approximate Pareto front on test function ZDT1 for an exemplary algorithm of the present invention;
FIG. 4 is a flow chart of an exemplary application of the large-scale multi-objective combinatorial optimization method provided by the present invention.
Detailed Description
The invention will now be described in detail with reference to the drawings and specific examples. The present embodiment is implemented on the premise of the technical scheme of the present invention, and a detailed implementation manner and a specific operation process are given, but the protection scope of the present invention is not limited to the following examples.
Example 1
According to the multi-objective optimization method provided by the invention, the problems of convergence and diversity are respectively processed in different optimization stages through complementary search strategies. In the first stage, the population approaches Pareto front rapidly, ignoring the diversity of the population. The Pareto optimal set (PS) is directly tracked by problem reconstruction when a selection pressure is encountered. The algorithm firstly acquires a group of reference directions in a decision space, associates the reference directions with a group of weight variables for positioning PS, and then converts the original large-scale multi-objective optimization problem into a low-dimensional single-objective optimization problem. In the second stage, the optimal solution is uniformly distributed on the approximate Pareto optimal front by using a multi-objective evolutionary algorithm. In the first stage, the population approaches the pareto front rapidly, ignoring the diversity of the population. And after the population converges, the decision variable clustering method is adopted in the second stage to emphasize the diversity of the population, so that the situation of sinking into local optimum is avoided.
As shown in fig. 1, a large-scale multi-objective combination optimization method based on problem recombination is shown, and the population is optimized in two stages in the framework: a Convergence-oriented Stage (CS) and a Diversity-oriented Stage (DS). The convergence optimization stage (CS) firstly adopts binary competition selection based on convergence degree, and uses crossover and mutation operators to select individuals from the current population P to generate N sub-generations Q, wherein N is the population scale. Then, the convergence of the union of P and Q is determined by equation (1). Finally, the optimal N individuals are selected from the union as new parent population P ′ When the population encounters the selection pressure, a group of weight vectors is maintained in a problem reconstruction mode, so that the population is large-scale and multi-scaleThe target optimization is converted into single target optimization, a child population A is produced through differential evolution, and the optimal N individuals are selected from the child population A and the weight vector population Q 'in a union mode based on an environment selection mechanism of a convergence degree strategy to serve as a new parent population P', so that the convergence of the population is further improved; the diversity optimization stage (DS) selects individuals from the current population P 'using a crowded distance-based binary race selection, and then uses the selected individuals to cross and mutate only diversity-related variables, generating offspring Q'. Finally, half of the population is selected from the union of the parent P 'and the offspring Q' as the new parent population according to the pareto dominance and crowding distance. The evolution steps of the specific framework are as follows:
randomly initializing a population P, wherein the population scale is N;
decision variables are divided into two classes (convergence and diversity variables) based on decision variable clustering techniques.
Using binary competition selection based on convergence, and using crossover and mutation operators to select individuals from a current population P to generate N sub-generations Q, wherein N is the population scale;
calculating the convergence degree of each individual of P and Q, and selecting the optimal N individuals from the union as a new parent population P'; the convergence Cd of the solution x is calculated as:
where M represents the target number. Convergence is defined as the sum of the target values of x. The smaller the value of Cd, the better the convergence of the solution x for minimizing MOP.
After the termination condition is satisfied, a problem-based reconstruction strategy is adopted from the current population P ′ R solutions are selected as the reference solution set. The two direction vectors for each reference solution are then calculated using equation 2.
Vl l ＝s 1 -o
Vu u ＝t-s 1 (2)
Wherein s is 1 ＝{x 1 ,…,x d Is } isReferring to the solution, o and t are the upper and lower boundary points of X, respectively.
Two weight vectors for each reference solution are calculated using equation 3.
Wherein lambda is 11 And lambda (lambda) 12 Is two weight variables, l max ＝||t-o||。
A set of reference solutions of size r is selected, and once each reference solution is associated with two direction vectors and two weight variables, a total of 2r sub-questions can be constructed.
The sub-problem is Z' (ζ) = { Z 11 (λ 11 )，z 12 (λ 12 )，…，z r1 (λ r1 )z r2 (λ r2 ) }, where ∈ = { λ 11 ，λ 12 ，…，λ r1 ，λ r2 The reconstructed decision space.
Once the sub-problem is reconstructed, the optimization of the decision vector x in the original decision space is translated into an optimization of the weight vector. Accordingly, the target space can be reduced, and the new optimization problem can be restated as
maximize G(∧)＝H(Z′(∧)) (5)
Where H may be any performance indicator.
Offspring population a was produced by differential evolution and the optimal N individuals from the union of a and Q' were selected as new parent population P "based on the environmental selection mechanism of the convergence strategy.
After the termination condition is satisfied, the DS is entered, first, individuals are selected from the current population P 'by adopting binary competition selection based on crowding distance, and then only the diversity related variables are crossed and mutated by the selected individuals, so that N offspring Q' are generated. Finally, according to the pareto dominance and crowding distance, half of the population is selected from the union of P 'and Q' to be used as a new parent population P ', after the termination condition is met, the parent population P' "is output as the optimal solution set for the optimization objective.
1) CS adopts directional cross variation facing convergence in the evolution process, and maintains the convergence of the population by combining with a convergence environment selection mechanism;
2) The DS adopts diversity-oriented directional cross variation in the evolution process, and maintains population diversity by combining a diversity environment selection mechanism.
Preferably, the above convergence and diversity crossover and mutation operators are implemented by convergence calculation and decision variable analysis, respectively, and the specific steps are as follows:
1) Decision variable clustering: the method adopts k-means clustering to divide variables into two classes (convergence variable and diversity variable).
2) Classification of convergence variable: the convergence Cd of the solution x is shown in the calculation formula (1).
3) The convergence crossover is performed by taking independent convergence variable groups as units; the diversity crossover is performed by taking all diversity variables as units;
4) The convergence variation is to select a variable from the convergence variables to vary; diversity variation among the diversity variations, the variable was selected for variation.
The problem reconstruction strategy used by CS in the evolution process is specifically as follows:
problem reconstruction strategy: this strategy is used to address the selection pressure caused by large-scale decision variables. Specifically, the method comprises three steps: the first step, weight vector association, selecting r solutions from the current population P as a reference solution set, wherein each reference solution is associated with two direction vectorsAnd two weight variables. Second, construct sub-problem Z' (a) = { Z according to formulas 4 and 5 11 (λ 11 )，z 12 (λ 12 )，…，z r1 (λ r1 )z r2 (λ r2 ) }, where ∈ = { λ 11 ，λ 12 ，…，λ r1 ，λ r2 The reconstructed decision space. Thirdly, reconstructing the target space, and once the sub-problem is reconstructed, converting the optimization of the decision vector x in the original decision space into the optimization of the weight vector lambda in the reconstructed decision space. The new optimization problem can be expressed as equation 5.
The diversity evolution stage can adopt other methods of multi-objective optimization algorithms, so that different instantiation algorithms are designed aiming at the multi-objective optimization problem of different types, and the expansibility is strong.
Table 1 shows the IGD mean values obtained over 9 large-scale test cases for two multi-objective evolutionary algorithms. To reduce the impact of random errors on the calculation results, each algorithm in this embodiment is independently run 30 times in each example, and the average value of the IGD index obtained by each algorithm in each example is calculated. The IGD index measures the distance between the real Pareto front to the algorithmically derived approximate Pareto. In general, the smaller the IGD index value, the better the convergence and diversity of the algorithm.
TABLE 1
| Problem | M | D | Algorithm for instantiation of the present framework | LSMOF |
| LSMOP1 | 2 | 1000 | 2.6390e-1(1.63e-2)+ | 6.2856e-1(2.77e-2) |
| LSMOP2 | 2 | 1000 | 1.1624e-2(3.74e-4)+ | 1.8858e-2(3.47e-4) |
| LSMOP3 | 2 | 1000 | 1.5680e+0(1.61e-3)+ | 1.5729e+0(4.09e-4) |
| LSMOP4 | 2 | 1000 | 2.5755e-2(7.90e-4)+ | 3.8112e-2(1.95e-3) |
| LSMOP5 | 2 | 1000 | 6.8558e-1(4.40e-2)+ | 7.4209e-1(3.39e-16) |
| LSMOP6 | 2 | 1000 | 5.6957e-1(1.62e-1)- | 3.1236e-1(4.03e-4) |
| LSMOP7 | 2 | 1000 | 1.5087e+0(4.30e-4)- | 1.5083e+0(7.70e-4) |
| LSMOP8 | 2 | 1000 | 5.3281e-1(2.17e-1)+ | 7.4209e-1(3.39e-16) |
| LSMOP9 | 2 | 1000 | 5.1013e-1(4.32e-3)+ | 8.0686e-1(8.94e-4) |
Fig. 2 is an approximate Pareto front obtained by the method of the present invention on a 2-target, 1000 decision variable LSMOP2 reference function, and fig. 3 is an approximate Pareto front obtained by the method of the present invention on a 2-target, 1000 decision variable ZDT1 reference function.
Example 2
As a second aspect of the present invention, there is provided an application of the method as described in the above embodiments, wherein a typical application scenario of the method is resource scheduling and integration optimization of a large-scale data center. The specific application process is as shown in fig. 4:
step 1: firstly, taking indexes such as energy consumption, resource loss, data communication flow, task completion time delay and the like as problem optimization targets, taking the residual available computing capacity of a server as constraint conditions, and establishing a multi-target constraint optimization model for large-scale data center resource scheduling and integration problems;
step 2: then, the number of virtual machines to be scheduled is taken as the chromosome length, and the number of the virtual machines is taken as the gene position, so that the coding of population individuals is realized;
step 3: finally, initializing a population P based on a coding mode, performing iterative optimization of the population according to a large-scale multi-objective combination optimization method based on problem recombination shown in fig. 1 until the termination condition of an algorithm is met, and outputting a Pareto optimal solution set taking a plurality of conflict indexes such as energy consumption, resource loss, data communication flow, task completion time delay and the like as optimization targets.
The foregoing describes in detail preferred embodiments of the present invention. It should be understood that numerous modifications and variations can be made in accordance with the concepts of the invention by one of ordinary skill in the art without undue burden. Therefore, all technical solutions which can be obtained by logic analysis, reasoning or limited experiments based on the prior art by the person skilled in the art according to the inventive concept shall be within the scope of protection defined by the claims.
Claims (10)
1. The large-scale multi-objective combination optimization method based on the problem recombination is characterized by comprising a convergence optimization stage and a diversity optimization stage, and comprises the following specific steps of:
randomly initializing a population P;
based on a decision variable clustering technology, dividing decision variables into convergence variables and diversity variables;
convergence optimization stage:
in the evolution process, adopting directional cross variation facing convergence, and combining with a convergence environment selection mechanism, selecting an optimal individual as a parent population P';
when the population encounters selection pressure, problem reconstruction is carried out, large-scale multi-objective optimization is converted into single-objective optimization, and the optimal individuals are reselected as parent populations P';
diversity optimization stage:
in the evolution process, adopting diversity-oriented directional cross variation, and combining a diversity environment selection mechanism to select a population as a new parent population;
and after the termination condition is met, outputting the parent population as an optimal solution set of the optimization target.
2. The problem recombination-based large-scale multi-objective combinatorial optimization method of claim 1, wherein the specific evolution steps of the convergence optimization stage are as follows:
using binary competition selection based on convergence, and selecting individuals from the current population to generate N sub-generations Q by using crossover and mutation operators;
calculating the convergence degree of each individual in the current population P and the offspring Q, and selecting the optimal N individuals from the union as a new parent population P';
repeating the above process until the termination condition is satisfied;
performing problem reconstruction, and converting large-scale multi-objective optimization into single-objective optimization through bi-directional weight vector association;
producing a progeny population a by differential evolution;
the convergence policy based environmental selection mechanism selects the optimal N individuals from the union of the offspring population a and the weight vector population Q' as the new parent population P ".
3. The method for large-scale multi-objective combinatorial optimization based on problem recombination according to claim 2, wherein,
the directional crossing of the convergence performs crossing operation by taking independent convergence variable groups as units;
the directional variation of the convergence selects a variable from the convergence variables to vary.
4. The problem recombination-based massive multi-objective combinatorial optimization method according to claim 2, wherein the convergence degree C of solution x d The calculation formula is as follows:
where M represents the target number.
5. The large-scale multi-objective combination optimization method based on problem recombination according to claim 1, wherein the problem recombination strategy is specifically: through bi-directional weight vector association, the original large-scale multi-objective optimization problem is re-expressed into single-objective optimization with relatively smaller weight variables, and a group of candidate solutions with good convergence and uniform distribution are used for guiding the algorithm to search the direction of the optimal set.
6. The method for large-scale multi-objective combinatorial optimization based on problem recombination according to claim 5, wherein the specific evolution steps of the problem recombination strategy are as follows:
weight vector correlation, selecting r solutions from the current population as a reference solution set, each reference solution being associated with two direction vectors V l And V u And two weight variables lambda r1 And lambda (lambda) r2 Associating;
constructing a sub-problem according to the direction vector and the weight variable:
Z′(∧)＝(z 11 (λ 11 )，z 12 (λ 12 )，...，z r1 (λ r1 )z r2 (λ r2 )}
wherein, lambada= { lambda 11 ，λ 12 ，...，λ r1 ，λ r2 A reconstructed decision space;
target space reconstruction, when the sub-problem is reconstructed, the optimization of the decision vector x in the original decision space is converted into the optimization of the weight vector Λ in the reconstructed decision space, and the new optimization problem is expressed as follows:
maximize G(Λ)＝H(Z′(Λ))
where H may be any performance indicator.
7. The method for massive multi-objective combinatorial optimization based on problem recombination according to claim 6, wherein the specific step of constructing the sub-problem according to the direction vector and the weight variable comprises:
two direction vectors for each reference solution are calculated:
V l ＝s 1 -o
V u ＝t-s 1
wherein s is 1 ＝{x 1 ，...，x d The } is a reference solution, and t and o are upper and lower boundary points of X respectively;
two weight vectors for each reference solution are calculated:
where λ11 and λ12 are two weight variables, l max ＝||t-o||；
The expression of the constructor problem is:
8. the problem recombination-based large-scale multi-objective combinatorial optimization method of claim 1, wherein the specific evolution steps of the diversity optimization stage are as follows:
selecting individuals from the current population P' using a binary competition selection based on crowding distances;
utilizing the selected individuals to only cross and mutate the diversity related variables to generate N offspring Q';
selecting half of the current population P ' and offspring Q ' from the union set as a new parent population P ' according to the pareto dominance and crowding distance;
repeating the above process until the termination condition is satisfied, and outputting the parent population P' as the optimal solution set of the optimization target.
9. The method for large-scale multi-objective combinatorial optimization based on problem recombination according to claim 8, wherein,
the directional crossing of the diversity performs crossing operation by taking all diversity variables as units;
the directed variation of diversity selects variables among the diversity variables for variation.
10. Application of the problem recombination-based large-scale multi-objective combination optimization method according to any of claims 1-9, characterized in that the application scenario of the method comprises resource scheduling and integration optimization of a large-scale data center; the specific application steps comprise:
taking indexes including energy consumption, resource loss, data communication flow and task completion time delay as problem optimization targets, taking the residual available computing capacity of a server as constraint conditions, and establishing a multi-target constraint optimization model of a large-scale data center resource scheduling and integration problem;
taking the number of virtual machines to be scheduled as the length of a chromosome, and taking the number of the virtual machines as the gene position, so as to realize the coding of population individuals;
initializing a population P based on a coding mode, performing iterative optimization of the population according to the large-scale multi-objective combination optimization method based on the problem recombination as claimed in any one of claims 1-9 until the termination condition of an algorithm is met, and outputting a Pareto optimal solution set with a plurality of conflict indexes including energy consumption, resource loss, data communication flow and task completion time delay as optimization objectives.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310233019.XA CN116306919A (en) | 2023-03-09 | 2023-03-09 | Large-scale multi-objective combination optimization method based on problem recombination and application |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310233019.XA CN116306919A (en) | 2023-03-09 | 2023-03-09 | Large-scale multi-objective combination optimization method based on problem recombination and application |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116306919A true CN116306919A (en) | 2023-06-23 |
Family
ID=86835528
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310233019.XA Pending CN116306919A (en) | 2023-03-09 | 2023-03-09 | Large-scale multi-objective combination optimization method based on problem recombination and application |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116306919A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117910410A (en) * | 2024-03-19 | 2024-04-19 | 电子科技大学 | Large-scale multi-target simulation chip circuit evolution optimization design method |
-
2023
- 2023-03-09 CN CN202310233019.XA patent/CN116306919A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117910410A (en) * | 2024-03-19 | 2024-04-19 | 电子科技大学 | Large-scale multi-target simulation chip circuit evolution optimization design method |
| CN117910410B (en) * | 2024-03-19 | 2024-05-31 | 电子科技大学 | Large-scale multi-target simulation chip circuit evolution optimization design method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Toutouh et al. | Spatial evolutionary generative adversarial networks | |
| CN102413029B (en) | Method for partitioning communities in complex dynamic network by virtue of multi-objective local search based on decomposition | |
| Wen et al. | Learning ensemble of decision trees through multifactorial genetic programming | |
| CN108470237B (en) | Multi-preference high-dimensional target optimization method based on co-evolution | |
| CN106991295B (en) | Protein network module mining method based on multi-objective optimization | |
| CN103745258A (en) | Minimal spanning tree-based clustering genetic algorithm complex web community mining method | |
| CN116306919A (en) | Large-scale multi-objective combination optimization method based on problem recombination and application | |
| CN111369000A (en) | High-dimensional multi-target evolution method based on decomposition | |
| CN111709526A (en) | Multi-mode and multi-target evolution algorithm based on multi-factor transfer learning | |
| Li et al. | A self‐organizing weighted optimization based framework for large‐scale multi‐objective optimization | |
| CN115481727A (en) | Intention recognition neural network generation and optimization method based on evolutionary computation | |
| CN107577918A (en) | The recognition methods of CpG islands, device based on genetic algorithm and hidden Markov model | |
| Zhu et al. | Gp-cnas: Convolutional neural network architecture search with genetic programming | |
| Broni-Bediako et al. | Evolutionary NAS with gene expression programming of cellular encoding | |
| Phan et al. | Efficiency enhancement of evolutionary neural architecture search via training-free initialization | |
| Ripon et al. | Evolutionary multi-objective clustering for overlapping clusters detection | |
| CN115293430A (en) | Unmanned node cooperation method and system based on cooperative coevolution algorithm | |
| CN116611504A (en) | Neural architecture searching method based on evolution | |
| CN114493047B (en) | Many-to-many goods scheduling optimization distribution method based on available resource coding | |
| CN114819181A (en) | Multi-target federal learning evolution method based on improved NSGA-III | |
| CN111883208B (en) | Gene sequence optimization method, device, equipment and medium | |
| Song et al. | Optimization algorithm of evolutionary design of circuits based on genetic algorithm | |
| Wei et al. | An improved genetic FCM clustering algorithm | |
| CN110059806B (en) | Multi-stage weighted network community structure detection method based on power law function | |
| CN113704570A (en) | Large-scale complex network community detection method based on self-supervision learning type evolution |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |