CN107977701A - A kind of online object space division methods, device and storage medium - Google Patents
A kind of online object space division methods, device and storage medium Download PDFInfo
- Publication number
- CN107977701A CN107977701A CN201710779955.5A CN201710779955A CN107977701A CN 107977701 A CN107977701 A CN 107977701A CN 201710779955 A CN201710779955 A CN 201710779955A CN 107977701 A CN107977701 A CN 107977701A
- Authority
- CN
- China
- Prior art keywords
- target
- population
- division
- object space
- space
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/086—Learning methods using evolutionary algorithms, e.g. genetic algorithms or genetic programming
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- General Physics & Mathematics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Physiology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention provides a kind of online object space division methods, device and storage medium, often take object space partitioning technology that object space is divided into some subspaces after an evolution generation, any target is not abandoned, so that each target is involved in evolutional operation, object space division i.e. of the invention is in line style, and it is to be based on collision probability information, object space can all be repartitioned in every generation of evolutionary computation, it is alternately that object space, which is divided with evolutional operation, avoid the problem of optimizing ability of prior art evolution algorithm is substantially obstructed, and increasing with target dimension, computation complexity will not rapidly increase.
Description
Technical field
The present invention relates to evolutionary computation technique field, more particularly to a kind of online object space division methods, device and deposit
Storage media.
Background technology
Based on the assumption that need the multi-objective optimization question optimized（Multi-objective optimization
Problems, MOPs）For the objective optimisation problems of minimum.Its mathematical description is：
Wherein, X is the feasible solution domain of problem, and x is the feasible solution vector of a V dimension,For problem
Goal set, M be problem target number, i.e. object space dimension,For scalar objective letter
Number, F (x) is the picture that feasible solution x is mapped in object space, it is a M dimensional vector function.When target number M is 2-3,
The problem is known as multi-objective optimization question, referred to as MOPs, if during target number M >=4, which is known as higher-dimension objective optimization
Problem（Many-objective optimization problems, MaOPs）, referred to as MaOPs.
Current multi-objective Evolutionary Algorithm（Multi-objective evolutionary algorithms, MOEAs）
（For example, NSGA-II【Deb K., Pratap A., Agarwal S., et al. A fast and elitist
multiobjective genetic algorithm: NSGA-II[J]. IEEE Transactions on
Evolutionary Computation, 2002, 6(2):182-197.】Deng classical multi-objective Evolutionary Algorithm）Can effectively it locate
The MOPs. that reason target number is 2-3 runs into understanding convergence difficulties when in face of the solution of MaOPs, computation complexity rapidly with
Target number increases and rapidly rises, and Pareto forward positions（Definition is see document【Giagkiozis, I. and P. J.
Fleming (2014). Pareto front estimation for decision making. Evolutionary
Computation, 22(4): pp 651-78.】）Visualization problem etc..Therefore, researcher is on the basis of general MOEAs
On propose many innovatory algorithms for MaOPs or introduce other tactful methods.One of research method is pair
The yojan of the target number of MaOPs, i.e. target dimensionality reduction, i.e., it is importance in MaOPs is low or the target identification of redundancy comes out, and
Abandoned, then remaining target is solved.But for the low target of importance, if being abandoned, gesture
The loss for the structural information that must throw into question, for the limitation of target dimensionality reduction, researcher proposes the side divided with object space
Method solves MaOPs, and researcher draws by experimental verification, and object space division methods solve MaOPs and can improve
The optimization performance of general MOEAs processing MaOPs.
Conventional object space division methods processing MaOPs be operated using MOEA in former problem object space it is near
Like sample data set of the disaggregation as division object space, object space is divided into several sub-goal spaces, and by population
（I.e. approximate disaggregation, is also referred to as Pareto approximation disaggregation in multiple-objection optimization field）Sub- population is divided into, every sub- population exists
Scanned in different sub-goal spaces, select the follow-on part of composition；Or population does not divide, but whole population
Scanned in different sub-goal spaces, select the follow-on part of composition.The common ground of the way of both is use
It is to evolve to obtain in former space in the sample data set of division object space, subspace is served only for the sequence of sub- population or population,
Parent is selected, evolutional operation is not carried out in subspace, is similarly to offline target dimension-reduction algorithm【Saxena, D. K.
and J. X. Duro, et al. (2013). Objective Reduction in Many-Objective
Optimization: Linear and Nonlinear Algorithms. IEEE Transactions on
Evolutionary Computation, 17(1): pp 77-99.】, i.e. the sample data set of dimensionality reduction is in former object space
The approximate disaggregation obtained after evolution algorithm is performed, and then is reduced the dimension of object space using dimension reduction method, obtains crucial mesh
Subset is marked, then population is sorted, and decision-making is selected to be carried out in common-denominator target subset.The key of this offline way is used
MOEAs is capable of providing the sample data set of good quality, otherwise can mislead decision-making, finally so that whole algorithm is in object space
Non-critical areas optimizing, so as to be unfavorable for correct decisions.Conventional object space division methods similar to offline target dimension reduction method,
Unlike, object space division methods processing higher-dimension objective optimization does not abandon target, but, all targets are involved in determining
Plan process, but according to the reality of Jaimes et al. the processing of the object space division methods based on goal conflict information MaOP proposed
Test conclusion, can object space affect algorithm performance by accurate division, this can be depended on by accurate division metric objective it
Between conflict degree method whether effectively or efficiently, they select Pearson correlations to measure goal conflict degree, and
And whether measurement is effectively decided by the quality of sample data set used, i.e., evolution algorithm used is after the execution of former object space
Whether the quality of the approximate disaggregation of gained is preferable.That is, either offline dimension-reduction algorithm or the division of offline target space
Method, their performance are all influenced be subject to the validity of evolution algorithm, and for higher-dimension objective optimisation problems, evolution algorithm
Optimizing ability（That is, the ability of Pareto approximation disaggregation is solved）Substantially it is obstructed, and increasing with target dimension, calculate multiple
Miscellaneous degree also rapidly increases.
Therefore, the prior art has yet to be improved and developed.
The content of the invention
In view of in place of above-mentioned the deficiencies in the prior art, it is an object of the invention to provide a kind of online object space division side
Method, device and storage medium, it is intended to offline dimension-reduction algorithm or offline target space-division method in the prior art is solved, they
Performance is all influenced be subject to the validity of evolution algorithm, and for higher-dimension objective optimisation problems, the optimizing ability of evolution algorithm
Substantially it is obstructed, and increasing with target dimension, computation complexity also burgeoning problem.
In order to achieve the above object, this invention takes following technical scheme：
A kind of online object space division methods, wherein, it the described method comprises the following steps：
S1, in solution space generate equally distributed initial population, obtains that corresponding with initial population, target is in whole two-by-two
Conflicting information matrix in solution space, carries out initial division to object space according to conflicting information matrix, obtains object space
First division；
S2, to carrying out independent evolution in each target subspace in first of object space division of initial population
Calculate, obtain part corresponding with each target subspace new explanation, and will part corresponding with each target subspace it is new
Solution is merged into the second population；
S3, judge whether the second population reaches pre-set end condition, when the second population reaches pre-set termination bar
Step S4 is then performed during part, step S5 is then performed when the second population is not up to pre-set end condition；
S4, using the second population export as Pareto optimality disaggregation；
S5, using the second population as initial population, return and perform step S1.
The online object space division methods, wherein, the step S1 is specifically included：
S11, in solution space generate equally distributed initial population, and initial population is denoted as A；
S12, carry out initial population A as sample two-by-two between target relative to the conflicting in whole solution space point
Analysis, obtains conflicting information matrix of the target in whole solution space two-by-two；
S13, according to conflicting information matrix carry out initial division to object space, first of object space division is obtained, by mesh
First division in mark space is denoted as { F1, F2 ..., Fn }；Wherein, n is obtained after the first division of object space progress
The number of target subspace, Fi is denoted as by target subspace, and 1≤i≤n, n are the positive integer more than or equal to 1.
The online object space division methods, wherein, the step S2 is specifically included：
S21, will be carried out successively in each sub-spaces of the initial population in first division { F1, F2 ..., Fn } it is non-dominant
Sequence, former generation's selection, after intersecting and making a variation, obtain part new explanation Ai corresponding with target subspace Fi, and Ai=MOEA（Fi）；
Wherein MOEA represents the multi-target evolution operator operated in target subspace Fi；
S22, merge part new explanation Ai corresponding with each target subspace Fi, obtains the second population, and by the second population
It is denoted as A '.
The online object space division methods, wherein, end condition is described in the step S3：Reach and preset
Evolutionary generation.
The online object space division methods, wherein, the part new explanation is N/n before being selected in target subspace
Solution is used as part new explanation；Wherein, N is the population scale of initial population, and n is obtained mesh after the first division of object space progress
Mark the number of subspace.
A kind of online object space division device, wherein, including：
Processor, is adapted for carrying out each instruction；And
Storage device, suitable for storing a plurality of instruction, described instruction is suitable for being loaded by processor and performing following steps：
Equally distributed initial population is generated in solution space, acquisition is corresponding with initial population, target is solved in whole two-by-two
Conflicting information matrix in space, carries out initial division to object space according to conflicting information matrix, obtains the of object space
One division；
To carrying out independent evolution meter in each target subspace of initial population in first division of object space
Calculate, obtain part corresponding with each target subspace new explanation, and will part corresponding with each target subspace new explanation
It is merged into the second population；
Judge whether the second population reaches pre-set end condition；
When the second population reaches pre-set end condition, then the second population is exported as Pareto optimality disaggregation,
When the second population is not up to pre-set end condition, then using the second population as initial population, returns to execution and obtain
, two-by-two target conflicting information matrix in whole solution space in corresponding with initial population is taken, according to conflicting information matrix to mesh
Mark space and carry out initial division, obtain first division of object space.
The online object space division device, wherein, it is described to generate equally distributed initial population in solution space,
, two-by-two target conflicting information matrix in whole solution space in corresponding with initial population is obtained, according to conflicting information matrix pair
Object space carries out initial division, and the step of obtaining first of object space division specifically includes：
Equally distributed initial population is generated in solution space, initial population is denoted as A；
Analyze, obtain relative to the conflicting in whole solution space using initial population A as between sample progress two-by-two target
To conflicting information matrix of the target two-by-two in whole solution space；
Initial division is carried out to object space according to conflicting information matrix, first division of object space is obtained, by target empty
Between first division be denoted as { F1, F2 ..., Fn }；Wherein, n is obtained target after the first division of object space progress
The number of subspace, Fi is denoted as by target subspace, and 1≤i≤n, n are the positive integer more than or equal to 1.
The online object space division device, wherein, it is described to initial population in first division of object space
Each target subspace in carry out independent evolutionary computation, it is new to obtain part corresponding with each target subspace
Solution, and will specifically include part corresponding with each target subspace new explanation the step of being merged into the second population：
Non-dominant row will be carried out successively in each sub-spaces of the initial population in first division { F1, F2 ..., Fn }
Sequence, former generation's selection, after intersecting and making a variation, obtain part new explanation Ai corresponding with target subspace Fi, and Ai=MOEA（Fi）；Its
Middle MOEA represents the multi-target evolution operator operated in target subspace Fi；
Part new explanation Ai corresponding with each target subspace Fi is merged, obtains the second population, and the second population is denoted as
A’。
The online object space division device, wherein, the end condition is：Reach evolutionary generation set in advance.
A kind of storage medium, wherein, wherein being stored with a plurality of instruction, described instruction is suitable for being loaded by processor and performing institute
The step of stating online object space division methods.
Beneficial effect：Online object space division methods, device and storage medium provided by the invention, often after an evolution generation
Take object space partitioning technology that object space is divided into some subspaces, do not abandon any target so that each mesh
Mark is involved in evolutional operation, i.e., object space of the invention division is in line style and is to be based on collision probability information, is being evolved
The every generation calculated can all repartition object space, and object space division and evolutional operation are alternately, to keep away
The problem of optimizing ability of prior art evolution algorithm is substantially obstructed, and increasing with target dimension are exempted from, have calculated complicated
Degree will not rapidly increase.
Brief description of the drawings
Fig. 1 is the flow chart of online object space division methods preferred embodiment of the present invention.
Fig. 2 is the flow chart of step S100 in online object space division methods of the present invention.
Fig. 3 is the flow chart of step S200 in online object space division methods of the present invention.
Embodiment
The present invention provides a kind of online object space division methods, device and storage medium, to make the purpose of the present invention, skill
Art scheme and effect are clearer, clear and definite, and the present invention is described in more detail for the embodiment that develops simultaneously referring to the drawings.It should manage
Solution, specific embodiment described herein only to explain the present invention, are not intended to limit the present invention.
Referring to Fig. 1, it is the flow chart of online object space division methods preferred embodiment of the present invention.Such as Fig. 1
Shown, the online object space division methods, comprise the following steps：
Step S1, generate equally distributed initial population in solution space, obtain it is corresponding with initial population, two-by-two target in
Conflicting information matrix in whole solution space, carries out initial division to object space according to conflicting information matrix, obtains target empty
Between first division；
Step S2, it is independent to being carried out in each target subspace in first of object space division of initial population
Evolutionary computation, obtains part corresponding with each target subspace new explanation, and will portion corresponding with each target subspace
New explanation is divided to be merged into the second population；
Step S3, judge whether the second population reaches pre-set end condition, when the second population reaches pre-set end
Step S4 is only then performed during condition, step S5 is then performed when the second population is not up to pre-set end condition；
Step S4, the second population is exported as Pareto optimality disaggregation；
Step S5, using the second population as initial population, return and perform step S1.
In this implementation, first, equally distributed one group of initial population is generated in solution space, since initial population is uniform
Be distributed in solution space, therefore, by the use of in initial population data as sample come the target two-by-two of problem analysis（It is every in MaOPs
Two targets）Between conflicting in whole solution space, so as to draw conflicting information of the target in whole solution space two-by-two
Matrix（Each element of matrix is the conflict degree between each two target）, then target empty is carried out according to this conflicting information matrix
Between initial division, obtain first of object space division（The division is made of F1, F2 ..., Fn）.Then, using general
Each target subspace Fi for being divided to initial population at this of MOEA in carry out independent evolutional operation, finally each
Selected section new explanation in target subspace, merges into the new population of algorithm（That is the second population）；Then before the second population is replaced
The initial population in face, repetitive operation（Object space is repartitioned using new population, and it is empty in the son of each division
Between in evolutional operation）, until meeting end condition（For example, reach evolutionary generation set in advance）Terminate, finally export the
Approximate Pareto optimality disaggregation of two populations as problem（In multi-objective optimization question, it is impossible to as single-object problem
It is equally obtaining the result is that unique, it is obtaining the result is that one group of solution, these are that all targets can connect in multi-objective problem
The satisfactory solution received, such one group of solution are also referred to as Pareto optimal solution sets in multiple target field）.
Online object space division methods of the present invention are dropped with being in the difference of line target dimension-reduction algorithm in line target
Dimension algorithm is to abandon redundancy object after evolving every time（Neglected, will not be caused damages to former structure of problem information
Target）, evolutional operation is then carried out on remaining common-denominator target.
Online object space division methods of the present invention and the difference of general object space partitioning technology are, conventional
Object space division methods processing higher-dimension objective optimisation problems（MaOP）It is to be grasped using evolution algorithm in former problem object space
Sample data set of the approximate disaggregation for making to obtain as division object space, several sub-goals sky is divided into by object space
Between, and be sub- population by population dividing, every sub- population scans in different sub-goal spaces, and it is of future generation to select composition
A part；Or population does not divide, but whole population scans in different sub-goal spaces, and it is next to select composition
The part in generation.Conventional object space division methods processing higher-dimension objective optimisation problems（MaOP）It is being total to for the way of both
With point to be to evolve to obtain in former space for the sample data set for dividing object space, subspace is served only for sub- population or population
Sequence, select parent, do not carry out evolutional operation in subspace, be similarly to offline target dimension-reduction algorithm, is i.e. dimensionality reduction
Sample data set is the approximate disaggregation obtained after former object space execution evolution algorithm, and then utilizes dimension reduction method by target empty
Between dimension reduce, obtain common-denominator target subset, then population sort, decision-making is selected to be carried out in common-denominator target subset.
And the present invention is to be based on collision probability information, object space partitioning technology often is taken by target empty after an evolution generation
Between be divided into some subspaces, do not abandon any target so that each target is involved in evolutional operation, i.e. target of the invention
Space division is in line style, namely the object space division methods based on collision probability information of the present invention are in evolutionary computation
Every generation object space can all be repartitioned, object space division with evolutional operation be alternately.
Preferably, as shown in Fig. 2, in the online object space division methods, the step S1 is specifically included：
Step S11, equally distributed initial population is generated in solution space, initial population is denoted as A；
Step S12, carried out initial population A as sample two-by-two between target relative to the conflicting in whole solution space
Analysis, obtains conflicting information matrix of the target in whole solution space two-by-two；
Step S13, initial division is carried out to object space according to conflicting information matrix, obtains first division of object space,
First division of object space is denoted as F1, F2 ..., Fn }；Wherein, n is gained after the first division of object space progress
The number of the target subspace arrived, Fi is denoted as by target subspace, and 1≤i≤n, n are the positive integer more than or equal to 1.
Preferably, as shown in figure 3, in the online object space division methods, the step S2 is specifically included：
Step S21, will be carried out successively in each sub-spaces of the initial population in first division { F1, F2 ..., Fn } non-
Dominated Sorting, former generation's selection, after intersecting and making a variation, obtain part new explanation Ai corresponding with target subspace Fi, and Ai=MOEA
（Fi）；Wherein MOEA represents the multi-target evolution operator operated in target subspace Fi；
Step S22, part new explanation Ai corresponding with each target subspace Fi is merged, obtains the second population, and by second
Population is denoted as A '.
Preferably, in the online object space division methods, end condition is described in the step S3：Reach pre-
The evolutionary generation first set.
Preferably, in the online object space division methods, the part new explanation is to be selected in target subspace
Preceding N/n solution is used as part new explanation；Wherein, N is the population scale of initial population, and n is institute after the first division of object space progress
The number of obtained target subspace.
Online object space division methods are based on based on above-mentioned, the present invention also provides a kind of online object space division dress
Put, wherein, including：
Processor, is adapted for carrying out each instruction；And
Storage device, suitable for storing a plurality of instruction, described instruction is suitable for being loaded by processor and performing following steps：
Equally distributed initial population is generated in solution space, acquisition is corresponding with initial population, target is solved in whole two-by-two
Conflicting information matrix in space, carries out initial division to object space according to conflicting information matrix, obtains the of object space
One division；
To carrying out independent evolution meter in each target subspace of initial population in first division of object space
Calculate, obtain part corresponding with each target subspace new explanation, and will part corresponding with each target subspace new explanation
It is merged into the second population；
Judge whether the second population reaches pre-set end condition；
When the second population reaches pre-set end condition, then the second population is exported as Pareto optimality disaggregation,
When the second population is not up to pre-set end condition, then using the second population as initial population, returns to execution and obtain
, two-by-two target conflicting information matrix in whole solution space in corresponding with initial population is taken, according to conflicting information matrix to mesh
Mark space and carry out initial division, obtain first division of object space.
Preferably, it is described equally distributed just in solution space generation in the online object space division device
Beginning population, obtains, two-by-two target conflicting information matrix in whole solution space in corresponding with initial population, according to conflicting information
Matrix carries out initial division to object space, and the step of obtaining first of object space division specifically includes：
Equally distributed initial population is generated in solution space, initial population is denoted as A；
Analyze, obtain relative to the conflicting in whole solution space using initial population A as between sample progress two-by-two target
To conflicting information matrix of the target two-by-two in whole solution space；
Initial division is carried out to object space according to conflicting information matrix, first division of object space is obtained, by target empty
Between first division be denoted as { F1, F2 ..., Fn }；Wherein, n is obtained target after the first division of object space progress
The number of subspace, Fi is denoted as by target subspace, and 1≤i≤n, n are the positive integer more than or equal to 1.
Preferably, in the online object space division device, it is described to initial population at first of object space
Independent evolutionary computation is carried out in each target subspace in division, obtains portion corresponding with each target subspace
Point new explanation, and will specifically include part corresponding with each target subspace new explanation the step of being merged into the second population：
Non-dominant row will be carried out successively in each sub-spaces of the initial population in first division { F1, F2 ..., Fn }
Sequence, former generation's selection, after intersecting and making a variation, obtain part new explanation Ai corresponding with target subspace Fi, and Ai=MOEA（Fi）；Its
Middle MOEA represents the multi-target evolution operator operated in target subspace Fi；
Part new explanation Ai corresponding with each target subspace Fi is merged, obtains the second population, and the second population is denoted as
A’。
Preferably, in the online object space division device, the end condition is：Reach evolution set in advance
Algebraically.
Device is divided based on online object space based on above-mentioned, the present invention also provides a kind of storage medium, wherein, wherein depositing
Contain a plurality of instruction, the step of described instruction is suitable for load by processor and being performed the object space division methods online.
In conclusion online object space division methods, device and storage medium provided by the present invention, a generation of often evolving
Afterwards take object space partitioning technology that object space is divided into some subspaces, do not abandon any target so that each
Target is involved in evolutional operation, i.e., object space of the invention division is in line style and is to be based on collision probability information, into
Change calculate every generation object space can all be repartitioned, object space division and evolutional operation be alternately,
The problem of optimizing ability of prior art evolution algorithm is substantially obstructed, and increasing with target dimension are avoided, is calculated multiple
Miscellaneous degree will not rapidly increase.
It is understood that for those of ordinary skills, can be with technique according to the invention scheme and this hair
Bright design is subject to equivalent substitution or change, and all these changes or replacement should all belong to the guarantor of appended claims of the invention
Protect scope.
Claims (10)
1. a kind of online object space division methods, it is characterised in that the described method comprises the following steps：
S1, in solution space generate equally distributed initial population, obtains that corresponding with initial population, target is in whole two-by-two
Conflicting information matrix in solution space, carries out initial division to object space according to conflicting information matrix, obtains object space
First division；
S2, to carrying out independent evolution in each target subspace in first of object space division of initial population
Calculate, obtain part corresponding with each target subspace new explanation, and will part corresponding with each target subspace it is new
Solution is merged into the second population；
S3, judge whether the second population reaches pre-set end condition, when the second population reaches pre-set termination bar
Step S4 is then performed during part, step S5 is then performed when the second population is not up to pre-set end condition；
S4, using the second population export as Pareto optimality disaggregation；
S5, using the second population as initial population, return and perform step S1.
2. online object space division methods according to claim 1, it is characterised in that the step S1 is specifically included：
S11, in solution space generate equally distributed initial population, and initial population is denoted as A；
S12, carry out initial population A as sample two-by-two between target relative to the conflicting in whole solution space point
Analysis, obtains conflicting information matrix of the target in whole solution space two-by-two；
S13, according to conflicting information matrix carry out initial division to object space, first of object space division is obtained, by mesh
First division in mark space is denoted as { F1, F2 ..., Fn }；Wherein, n is obtained after the first division of object space progress
The number of target subspace, Fi is denoted as by target subspace, and 1≤i≤n, n are the positive integer more than or equal to 1.
3. online object space division methods according to claim 2, it is characterised in that the step S2 is specifically included：
S21, will be carried out successively in each sub-spaces of the initial population in first division { F1, F2 ..., Fn } it is non-dominant
Sequence, former generation's selection, after intersecting and making a variation, obtain part new explanation Ai corresponding with target subspace Fi, and Ai=MOEA（Fi）；
Wherein MOEA represents the multi-target evolution operator operated in target subspace Fi；
S22, merge part new explanation Ai corresponding with each target subspace Fi, obtains the second population, and by the second population
It is denoted as A '.
4. online object space division methods according to claim 3, it is characterised in that terminate bar described in the step S3
Part is：Reach evolutionary generation set in advance.
5. according to any one of the claim 1-4 online object space division methods, it is characterised in that the part new explanation is
N/n solution is used as part new explanation before being selected in target subspace；Wherein, N is the population scale of initial population, and n is target empty
Between carry out first division after obtained target subspace number.
A kind of 6. online object space division device, it is characterised in that including：
Processor, is adapted for carrying out each instruction；And
Storage device, suitable for storing a plurality of instruction, described instruction is suitable for being loaded by processor and performing following steps：
Equally distributed initial population is generated in solution space, acquisition is corresponding with initial population, target is solved in whole two-by-two
Conflicting information matrix in space, carries out initial division to object space according to conflicting information matrix, obtains the of object space
One division；
To carrying out independent evolution meter in each target subspace of initial population in first division of object space
Calculate, obtain part corresponding with each target subspace new explanation, and will part corresponding with each target subspace new explanation
It is merged into the second population；
Judge whether the second population reaches pre-set end condition；
When the second population reaches pre-set end condition, then the second population is exported as Pareto optimality disaggregation,
When the second population is not up to pre-set end condition, then using the second population as initial population, returns to execution and obtain
, two-by-two target conflicting information matrix in whole solution space in corresponding with initial population is taken, according to conflicting information matrix to mesh
Mark space and carry out initial division, obtain first division of object space.
7. online object space division device according to claim 6, it is characterised in that described equal in solution space generation
The initial population of even distribution, obtains, two-by-two target conflicting information matrix in whole solution space in corresponding with initial population, root
The step of carrying out initial division according to conflicting information matrix to object space, obtain first division of object space specifically includes：
Equally distributed initial population is generated in solution space, initial population is denoted as A；
Analyze, obtain relative to the conflicting in whole solution space using initial population A as between sample progress two-by-two target
To conflicting information matrix of the target two-by-two in whole solution space；
Initial division is carried out to object space according to conflicting information matrix, first division of object space is obtained, by target empty
Between first division be denoted as { F1, F2 ..., Fn }；Wherein, n is obtained target after the first division of object space progress
The number of subspace, Fi is denoted as by target subspace, and 1≤i≤n, n are the positive integer more than or equal to 1.
8. online object space division device according to claim 7, it is characterised in that it is described to initial population in target empty
Between first division in each target subspace in carry out independent evolutionary computation, obtain and each target be empty
Between corresponding part new explanation, and will part corresponding with each target subspace new explanation the step of being merged into the second population it is specific
Including：
Non-dominant row will be carried out successively in each sub-spaces of the initial population in first division { F1, F2 ..., Fn }
Sequence, former generation's selection, after intersecting and making a variation, obtain part new explanation Ai corresponding with target subspace Fi, and Ai=MOEA（Fi）；Its
Middle MOEA represents the multi-target evolution operator operated in target subspace Fi；
Part new explanation Ai corresponding with each target subspace Fi is merged, obtains the second population, and the second population is denoted as
A’。
9. online object space division device according to claim 8, it is characterised in that the end condition is：Reach pre-
The evolutionary generation first set.
10. a kind of storage medium, it is characterised in that be wherein stored with a plurality of instruction, described instruction is suitable for by processor loading simultaneously
The step of any one of perform claim requirement 1-5 online object space division methods.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710779955.5A CN107977701A (en) | 2017-09-01 | 2017-09-01 | A kind of online object space division methods, device and storage medium |
| PCT/CN2017/119551 WO2019041700A1 (en) | 2017-09-01 | 2017-12-28 | Online target space dividing method and device, and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710779955.5A CN107977701A (en) | 2017-09-01 | 2017-09-01 | A kind of online object space division methods, device and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107977701A true CN107977701A (en) | 2018-05-01 |
Family
ID=62012380
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710779955.5A Pending CN107977701A (en) | 2017-09-01 | 2017-09-01 | A kind of online object space division methods, device and storage medium |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN107977701A (en) |
| WO (1) | WO2019041700A1 (en) |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103049800B (en) * | 2012-12-17 | 2016-08-03 | 上海大学 | A kind of automatic stereowarehouse scheduling Multipurpose Optimal Method storing the limited time |
| CN105719064A (en) * | 2016-01-15 | 2016-06-29 | 深圳大学 | Method and system for evaluating confliction degree between targets |
| CN106997486A (en) * | 2017-04-01 | 2017-08-01 | 中山大学 | A kind of higher-dimension multi-objective optimization algorithm split based on space |
| CN107092977A (en) * | 2017-04-01 | 2017-08-25 | 中山大学 | A kind of solution algorithm of the multiple target with time window isomery vehicle Location-Routing Problem |
-
2017
- 2017-09-01 CN CN201710779955.5A patent/CN107977701A/en active Pending
- 2017-12-28 WO PCT/CN2017/119551 patent/WO2019041700A1/en active Application Filing
Also Published As
| Publication number | Publication date |
|---|---|
| WO2019041700A1 (en) | 2019-03-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Xu et al. | Prediction-guided multi-objective reinforcement learning for continuous robot control | |
| CN107169768B (en) | Method and device for acquiring abnormal transaction data | |
| Sato et al. | Variable space diversity, crossover and mutation in MOEA solving many-objective knapsack problems | |
| CN110097088A (en) | A kind of dynamic multi-objective evolvement method based on transfer learning Yu particular point strategy | |
| Cai et al. | A hybrid of decomposition and domination based evolutionary algorithm for multi-objective software next release problem | |
| CN115277354B (en) | Fault detection method for command control network management system | |
| Pourchot et al. | To share or not to share: A comprehensive appraisal of weight-sharing | |
| Liu et al. | A decomposition-based evolutionary algorithm with correlative selection mechanism for many-objective optimization | |
| CN109767034B (en) | Relay protection constant value optimization method and device, computer equipment and storage medium | |
| Ripon et al. | Dynamic facility layout problem under uncertainty: a Pareto-optimality based multi-objective evolutionary approach | |
| Chen et al. | A modified adaptive switching-based many-objective evolutionary algorithm for distributed heterogeneous flowshop scheduling with lot-streaming | |
| Zhang et al. | Offline approximate value iteration for dynamic solutions to the multivehicle routing problem with stochastic demand | |
| CN109389517B (en) | Analysis method and device for quantifying line loss influence factors | |
| CN107977701A (en) | A kind of online object space division methods, device and storage medium | |
| Purshouse et al. | An adaptive divide-and-conquer methodology for evolutionary multi-criterion optimisation | |
| Kuk et al. | Combining fitness landscape analysis and adaptive operator selection in multi and many-objective optimization | |
| CN115619028A (en) | Clustering algorithm fusion-based power load accurate prediction method | |
| CN114117917B (en) | Multi-objective optimization ship magnetic dipole array modeling method | |
| CN115730432A (en) | Scheduling method, system, equipment and storage medium for data processing tasks of Internet of things | |
| Kostadinova et al. | An Integrative DTW-based imputation method for gene expression time series data | |
| CN107665291A (en) | A kind of mutation detection method based on cloud computing platform Spark | |
| Denysiuk et al. | A new hybrid evolutionary multiobjective algorithm guided by descent directions | |
| CN111092425B (en) | Net rack clustering analysis method and system based on topological characteristics | |
| CN110020725B (en) | Test design method for weapon equipment system combat simulation | |
| CN111383052A (en) | Intelligent cabinet site selection model modeling method and device, server and storage medium |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180501 |
|
| RJ01 | Rejection of invention patent application after publication |