CN114444645A - Intelligent optimization algorithm for multi-strategy multi-element universe group - Google Patents

Intelligent optimization algorithm for multi-strategy multi-element universe group Download PDF

Info

Publication number
CN114444645A
CN114444645A CN202210036297.1A CN202210036297A CN114444645A CN 114444645 A CN114444645 A CN 114444645A CN 202210036297 A CN202210036297 A CN 202210036297A CN 114444645 A CN114444645 A CN 114444645A
Authority
CN
China
Prior art keywords
universe
population
cosmic
matrix
fitness
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
Application number
CN202210036297.1A
Other languages
Chinese (zh)
Inventor
毕敬
舒文超
乔俊飞
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing University of Technology
Original Assignee
Beijing University of Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beijing University of Technology filed Critical Beijing University of Technology
Priority to CN202210036297.1A priority Critical patent/CN114444645A/en
Publication of CN114444645A publication Critical patent/CN114444645A/en
Pending legal-status Critical Current

Links

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
    • G06N3/004—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]

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention provides an improved intelligent optimization algorithm for a multi-strategy multi-universe group. The method improves and innovates the defects of single updating strategy, insufficient global exploration range and the like of the multi-element universe intelligent optimization algorithm, and expands the global exploration range by using the crossing, variation and selection mechanisms in the genetic algorithm for reference in the optimization process; and a space group updating strategy is innovated, an updating mechanism in the optimization process is enriched, and the local detection capability is enhanced. The specific operation is as follows: firstly, initializing a problem to be optimized, a population U to be optimized and a fitness function; on the basis, updating the cosmic population E to be optimized by using different strategies according to the initialization information; thirdly, reversely using different strategies to the updated E universe population to update the U universe population to be optimized; and finally, continuously iterating the forward and reverse optimization process to obtain a final optimization result.

Description

Intelligent optimization algorithm for multi-strategy multi-element universe group
Technical Field
The invention relates to the technical field of intelligent optimization algorithms, in particular to an intelligent optimization algorithm (MS _ MVO) of a multi-strategy multi-element universe group.
Background
The modern optimization algorithm is also called as an intelligent optimization algorithm or a modern heuristic algorithm, and is an algorithm which has global optimization performance and strong universality and is suitable for parallel processing. It has good effect in solving the optimization problem, and the inspiration mostly comes from different organisms in nature. The optimization problem refers to finding the optimal solution meeting the requirement under a certain constraint condition, and is a typical NP-hard problem. The most common in real life is the discrete combinatorial optimization problem, such as the traveler problem, the knapsack problem, the map coloring problem, the vehicle path problem, etc.
In the past, a large number of intelligent optimization algorithms have been proposed, typically ant colony optimization algorithms, particle swarm optimization algorithms, artificial bee colony algorithms, firefly algorithms, and the like. The theoretical basis of the algorithm is strict, and the optimal solution or the approximate optimal solution can be found in a certain time theoretically by only depending on related experience. Their common features are: the optimal solution is searched in the whole solution space with a certain probability according to a certain mechanism from any group of solutions. However, in the optimization process of the algorithms, the algorithms are likely to fall into a local optimal solution due to limited exploration space, so that the later convergence accuracy is unstable.
In the prior art, for example, a paper "Multi-version Optimizer: nature-embedded algorithm for globalization" published in 22.2.2015, the idea of Multi-universe is applied to the field of optimization algorithm, and different strategies are adopted for material exchange according to the properties (black holes, white holes, wormholes) of different universes to reduce the possibility of trapping in a local optimal solution, improve the global search capability of the algorithm, and effectively avoid premature convergence, but the exchange strategy is simpler and the local search capability is insufficient, resulting in poor final optimization effect.
Disclosure of Invention
The invention provides an intelligent optimization algorithm of a multi-strategy multi-element universe group, aiming at overcoming the technical defects of insufficient local search energy and insufficient material exchange strategy of the existing intelligent optimization algorithm. The method comprehensively considers and uses the advantages of various common intelligent optimization algorithms, thereby realizing the rapid excellence finding of the optimization problem.
In order to achieve the purpose, the intelligent optimization algorithm of the multi-strategy multi-element universe group adopts the following technical scheme:
step S1: setting the number of initial random universes (a plurality of solutions of a problem to be optimized) as m and the number of universe substances (parameters of the problem to be optimized) as n, establishing a universe population matrix U to be optimized in m rows and n columns, and initializing a corresponding universe population matrix E according to the initialized universe population matrix U;
step (ii) ofS2: determining an objective function and a fitness function of a solution, and calculating the fitness f (U) of the universe population U to be optimizedi) And fitness norm _ f (U) after normalizationi) And according to the fitness f (U)i) Finding out global optimal cosmic individuals global _ best of the population and historical optimal cosmic best of each cosmic individual in the populationiWhere f is the fitness function, UiThe ith universe in the universe population;
step S3: according to global _ best and bestiUpdating the E universe population matrix according to a certain rule;
step S4: according to the universe individual UiInitializing a new universe individual matrix OiAnd according to global _ best and bestiUpdating the data according to a certain strategy;
step S5: using cosmic object O updated in step S4iUpdating individual E in corresponding E universe populationiAnd record EiIf the iteration times are not changed, judging whether the current iteration E universe individual exceeds a specified number of times and is not updated (if the iteration times exceed the specified number, the current iteration E universe individual is likely to fall into a local optimal solution), if so, screening, covering and updating according to a certain strategy, otherwise, not changing;
step S6: judging whether all the individuals in the current E universe population are completely updated;
if all the updates are completed, the process proceeds to step S7;
if not, returning to the step S4 to continue to carry out individual iteration to update the E universe population;
step S7: calculating and sequencing fitness of all E universe populations updated above for updating universe population U to be optimized;
step S8: for the sorted E universe population, the universe individual E is selected by using a roulette mechanismindexAnd make EindexUpdating the history information of the universe individual and the universe population U to be optimized according to a certain strategyi；
Step S9: judging whether all the individuals in the U universe population are completely updated, if all the individuals in the U universe population are completely updated, entering the next step, and if not, returning to the step S8 to continue the iteration of the universe individuals;
step S10: judging whether a termination condition is reached;
if so, outputting the universe population with the optimal fitness as an optimal value to complete optimization calculation;
and if not, taking the new population matrix as the population matrix to be optimized, and returning to the step S2 to continue iterative optimization.
In the scheme, the E universe population is generated firstly, and is updated by using various strategies through iteration, so that the global exploration space is fully expanded; and then, directly performing iterative updating on the cosmic population to be optimized by using different strategies according to the updated E cosmic population and the historical information of the cosmic population U to be optimized. Therefore, the global search range is ensured, and the capability of local search is enhanced.
Preferably, the cosmic population matrix U to be optimized is initialized in step S1 using the Numpy package of Python according to the following code:
universes＝numpy.zeros((universe_count,universe_dim))
universes[:,i]＝numpy.random.uniform(0,1,m)*(up_list[i]-low_list[i])+low_list[i]
wherein the code represents a matrix of m rows and n columns, and each element is limited within a certain range; up _ list [ i ] and low _ list [ i ] are respectively the upper limit and the lower limit of the ith parameter corresponding to the problem to be optimized; the universe _ count represents the quantity of the universe population, and the universe _ dim represents the dimension of the problem to be optimized.
Preferably, in step S2, the ordered universe group sort _ U is sort (U, F):
wherein, F represents a matrix with m rows and 1 columns of each universe individual fitness, sort (U, F) represents that the fitness matrix is arranged according to the sequence from good to bad, the individuals in the universe group matrix U to be optimized are adjusted according to the sequence from good to bad of the fitness value, the adjusted matrix is output as an ordered universe group matrix sort _ U, and then according to the adjusted ordered universe matrix, global _ best and best are selectedi。
Preferably, the universe matrix E is updated at step S3 by the following formula:
Figure BDA0003464301800000021
wherein, c1、r1,d、c2、r2,dAre all in [0,1 ]]Uniformly selecting random numbers;
preferably, the cosmic matrix O is initialized and updated at step S4 by the following formula:
Figure BDA0003464301800000031
wherein r isdIs one in [0,1 ]]In a uniformly distributed random number, kdE {1,2, …, M } is an index of a random universe of individuals, pmIs a mutation probability to limit the probability of substance change in the universe individual, and is tested by experiment pmThe experimental effect is better when the value is 0.1; lbdAnd ubdRespectively, a minimum value and a maximum value of each element;
preferably, cosmic individuals E are updated at step S5 by the following strategyi：
Figure BDA0003464301800000032
If the cosmic object O generated in step S4 is presentiIf the fitness is better than the current universe individual EiThen use OiIn place of EiAnd zero setting the iteration times of the cosmic individuals without change; otherwise, the updating is not carried out, and the iteration times of the cosmic individuals without change are increased by one. Wherein the unchanged iteration number indicates that the current universe individual EiAnd the updating times are not generated in multiple iterations, and a certain cosmic individual can be effectively prevented from being trapped into a local optimal solution by limiting the unchanged iteration times.
For updated cosmic individuals EiIf the number of times of change exceeds the threshold value, 20% of cosmic individuals are selected from the E cosmic population, and the cosmic individuals are selected as the appropriate individualsReplacing the universe individual E to be updated with the universe individual with the best responsei(ii) a Otherwise, no change is made.
Preferably, the step S6 further includes the steps of:
after updating each universe individual EiIf the individuals which are not updated still exist, returning to the step S4 to update the new E cosmic population individuals; otherwise, the next step is carried out.
Preferably, in step S7, the ordered E universe population sort _ E is sort (E, F _ E):
wherein, F _ E represents a matrix of m rows and 1 columns of the fitness of each individual in the E cosmic population updated in step S6, and sort (U, F) represents that the fitness matrices are arranged in the order from good to bad, the individuals in the E cosmic population are adjusted in the order from good to bad according to the fitness values, and the adjusted matrix is output as an ordered cosmic population matrix sort _ E.
Preferably, the step S8 further includes the following steps to update the cosmic population U to be optimized:
first, designing the roulette mechanism to select E for updating the universe population UindexUniverse individuals:
firstly, carrying out cumulative summation on the sequential E universe population fitness according to the following formula;
Figure BDA0003464301800000033
then, [0, cumsum ] is generatedm]A random number p of the interval;
judging the fitness accumulation and the cumsum of each universe individualiAnd if the index is larger than p, returning the current index which is the E universe individual number selected by the roulette mechanism.
Secondly, designing an updating strategy of the universe population U:
Figure BDA0003464301800000034
Figure BDA0003464301800000035
wherein r2-r4 are in the number of [0,1 ]]In a uniformly distributed random number, norm _ f (U)i) Representing the universe individual UiThe normalized value of fitness, index, is the number of individuals in the E universe selected by roulette: TDR and WEP are both coefficients:
Figure BDA0003464301800000041
Figure BDA0003464301800000042
wherein p is the development progress in the universe iteration process, and the larger p is, the higher the exploration speed and the local optimization speed are, the more accurate p is (the algorithm tests show that p is 6, the better effect is obtained); min and max are fixed values, and the WEP value range is specified (the algorithm test finds that the min is 0.2, and the max is 1, so that the effect is better);
preferably, the termination condition in step S10 is either one of the following two conditions:
the first condition is as follows: the difference value between the optimal universe individual fitness output by the current iteration and the optimal universe individual fitness output by the last iteration is smaller than a set threshold;
and a second condition: the iteration optimization times reach the preset maximum iteration times.
The invention has the following beneficial effects:
the invention provides an intelligent optimization algorithm of a multi-strategy multi-element universe group, which is based on a plurality of common intelligent optimization algorithms, comprehensively considers the advantages and the defects of each algorithm and organically combines and innovates the algorithms. The E universe population is initialized and updated, the global exploration range is expanded, the E universe population and the information of the universe population to be optimized are used for optimizing the population to be optimized, and the local search space is deepened. Through the steps, the optimization searching process not only increases the global exploration range, but also avoids falling into local optimization; but also can go deep into the local search range and improve the optimization effect.
Drawings
The following detailed description of embodiments of the invention is provided in conjunction with the appended drawings:
FIG. 1 is a conceptual framework diagram of a multi-strategy intelligent optimization algorithm;
FIG. 2 is a flow chart of the implementation steps of the multi-strategy universe optimization algorithm;
FIG. 3 is a diagram illustrating the technical solution of the present invention in terms of the Benchmark function
Figure BDA0003464301800000043
Initializing a universe population position graph during solving;
FIG. 4 is a diagram illustrating the technical solution of the present invention in terms of the Benchmark function
Figure BDA0003464301800000044
Iteratively optimizing the result for 100 times during solving;
FIG. 5 is a diagram of a method for providing intelligent optimization and an original multi-element universe intelligent optimization algorithm in a Benchmark function
Figure BDA0003464301800000045
The contrast effect of (1);
FIG. 6 is a diagram of a method for providing intelligent optimization and an original multi-element universe intelligent optimization algorithm in a Benchmark function
Figure BDA0003464301800000046
The above contrast effect.
Detailed Description
To more clearly illustrate the intelligent optimization algorithm, the present invention is further described below with reference to preferred embodiments and the accompanying drawings. Similar parts in the figures are denoted by the same reference numerals. It is to be understood by persons skilled in the art that the following detailed description is illustrative and not restrictive, and is not to be taken as limiting the scope of the invention.
The invention provides an intelligent optimization algorithm of a multi-strategy multi-element universe group, which is characterized in that advantages and defects of the algorithm are obtained by analyzing various common intelligent optimization algorithms, and the algorithm is innovatively fused, so that the overall exploration space of the algorithm is increased, and the local optimization exploration and solution of the algorithm are deepened. As shown in fig. 2, the intelligent optimization algorithm includes the following steps:
step S1: setting the number of initial random universes (a plurality of solutions of a problem to be optimized) as m and the number of universe substances (parameters of the problem to be optimized) as n, establishing universe population matrixes U to be optimized in m rows and n columns, and initializing corresponding universe population matrixes E according to the initialized universe population matrixes U;
step S2: determining an objective function and a fitness function of a solution, and calculating the fitness f (U) of the universe population U to be optimizedi) And fitness norm _ f (U) after normalizationi) And according to the fitness f (U)i) Finding out global optimal cosmic individuals global _ best of the population and historical optimal cosmic best of each cosmic individual in the populationiWhere f is the fitness function, UiThe ith universe in the universe population;
step S3: according to global _ best and bestiUpdating the E universe population matrix according to a certain rule;
step S4: according to the universe individual UiInitializing a new universe individual matrix OiAnd according to global _ best and bestiUpdating the data according to a certain strategy;
step S5: using cosmic object O updated in step S4iUpdating individual E in corresponding E universe populationiAnd record EiIf the iteration times are not changed, judging whether the current iteration E universe individual exceeds a specified number of times and is not updated (if the iteration times exceed the specified number, the current iteration E universe individual is likely to fall into a local optimal solution), if so, screening, covering and updating according to a certain strategy, otherwise, not changing;
step S6: judging whether all the individuals in the current E universe population are completely updated;
if all the updates are completed, the process proceeds to step S7;
if not, returning to the step S4 to continue to carry out individual iteration to update the E universe population;
step S7: calculating and sequencing fitness of all E universe populations updated above for updating universe population U to be optimized;
step S8: for the sorted E universe population, the universe individual E is selected by using a roulette mechanismindexAnd make EindexUpdating the history information of the universe individual and the universe population U to be optimized according to a certain strategyi；
Step S9: judging whether all the individuals in the U universe population are completely updated, if all the individuals in the U universe population are completely updated, entering the next step, and if not, returning to the step S8 to continue the iteration of the universe individuals;
step S10: judging whether a termination condition is reached;
if so, outputting the universe population with the optimal fitness as an optimal value to complete optimization calculation;
and if not, taking the new population matrix as the population matrix to be optimized, and returning to the step S2 to continue iterative optimization.
In a specific implementation process, a m-row n-column cosmic population matrix U to be optimized is established, and a cosmic population E of the cosmic population U to be optimized is updated; firstly, updating the universe population E and sequencing according to fitness; and then updating the cosmic population U to be optimized by using the cosmic population E and the history information of the cosmic population U to be optimized. Wherein, steps S3 to S9 are processes of iterative optimization calculation.
More specifically, the cosmic population matrix U to be optimized is initialized in step S1 using the Numpy package of Python according to the following code:
universes＝numpy.zeros((universe_count,universe_dim))
universes[:,i]＝numpy.random.uniform(0,1,m)*(up_list[i]-low_list[i])+low_list[i]；
wherein the code represents a matrix of m rows and n columns, and each element is limited within a certain range; up _ list [ i ] and low _ list [ i ] are respectively the upper limit and the lower limit of the ith parameter corresponding to the problem to be optimized; the universe _ count represents the quantity of the universe population, and the universe _ dim represents the dimension of the problem to be optimized.
More specifically, in step S2, the ordered universe group sort _ u ═ Sort (U, F). Wherein, F represents a matrix with m rows and 1 columns of each universe individual fitness, sort (U, F) represents that the fitness matrix is arranged according to the sequence from good to bad, the individuals in the universe group matrix U to be optimized are adjusted according to the sequence from good to bad of the fitness value, the adjusted matrix is output as an ordered universe group matrix sort _ U, and then according to the adjusted ordered universe matrix, global _ best and best are selectedi。
More specifically, the cosmic matrix E is updated at step S3 by the following formula:
Figure BDA0003464301800000061
wherein, c1,r1,d,c2,r2,dAre all in [0,1 ]]Uniformly selecting random numbers;
preferably, the cosmic matrix O is initialized and updated at step S4 by the following formula:
Figure BDA0003464301800000062
Oi,d＝rand(lbd,ubd),if rd＜pm，
wherein r isdIs one in [0,1 ]]In a uniformly distributed random number, kdE {1, 2.., M } is an index of a random universe of individuals, pmIs a mutation probability to limit the probability of substance change in the universe individual, and is tested by experiment pmThe experimental effect is better when the value is 0.1; lbdAnd ubdRespectively, a minimum value and a maximum value of each element;
more specifically, step S5 further includes the following steps to update cosmic individuals Ei：
Figure BDA0003464301800000063
If the cosmic object O generated in step S4 is presentiIf the fitness is better than the current universe individual EiThen use OiIn place of EiAnd zero setting the iteration times of the cosmic individuals without change; otherwise, the updating is not carried out, and the iteration times of the cosmic individuals without changes are increased by one. Wherein the unchanged iteration number indicates that the current universe individual EiAnd the updating times are not generated in multiple iterations, and a certain cosmic individual can be effectively prevented from being trapped into a local optimal solution by limiting the unchanged iteration times.
For updated cosmic individuals EiIf the number of times of change exceeds the threshold value of unchanged times, selecting 20% of cosmic individuals from the E cosmic population, selecting the cosmic individuals with the best fitness from the E cosmic population, and then replacing the E cosmic individuals to be updatedi(ii) a Otherwise, no change is made.
More specifically, step S6 further includes the following steps:
after updating each universe individual EiIf the individuals which are not updated still exist, returning to the step S4 to update the new E cosmic population individuals; otherwise, the next step is carried out.
More specifically, in step S7, ordered E universe population sort _ E ═ sort (E, F _ E):
wherein, F _ E represents a matrix of m rows and 1 columns of the fitness of each individual in the E cosmic population updated in step S6, and sort (U, F) represents that the fitness matrices are arranged in the order from good to bad, the individuals in the E cosmic population are adjusted in the order from good to bad according to the fitness values, and the adjusted matrix is output as an ordered cosmic population matrix sort _ E.
More specifically, step S8 further includes the following steps to update the cosmic population U:
first, a roulette mechanism is designed to select E for updating universe population UindexUniverse individuals:
firstly, carrying out cumulative summation on the sequential E universe population fitness according to the following formula;
Figure BDA0003464301800000064
then, [0, cumsum ] is generatedm]A random number p of the interval; judging the fitness accumulation and the cumsum of each universe individualiAnd if the index is larger than p, returning the current index which is the E universe individual number selected by the roulette mechanism.
Secondly, designing an updating strategy of the universe population U:
Figure BDA0003464301800000071
Figure BDA0003464301800000072
wherein r2-r4 are in the number of [0,1 ]]In a uniformly distributed random number, norm _ f (U)i) Representing the universe individual UiThe normalized value of fitness, index, is the number of individuals in the E universe selected by roulette: TDR and WEP are both coefficients:
Figure BDA0003464301800000073
Figure BDA0003464301800000074
wherein p is the development progress in the universe iteration process, and the larger p is, the higher the exploration speed and the local optimization speed are, the more accurate p is (the algorithm tests show that p is 6, the better effect is obtained); min and max are fixed values, and the WEP value range is specified (the algorithm test shows that min is 0.2, and max is 1, so that the effect is better);
more specifically, the termination condition in step S10 is either one of the following two conditions:
the first condition is as follows: the difference value between the optimal universe individual fitness output by the current iteration and the optimal universe individual fitness output by the last iteration is smaller than a set threshold;
and a second condition: the iteration optimization times reach the preset maximum iteration times.
In this experiment, for the Benchmark function
Figure BDA0003464301800000075
In the iterative optimization process using the above algorithm, the result of the 100 th iteration is shown in fig. 4, which shows that the result is much better than that of the initialization, and the cosmic population to be optimized is more concentrated around the optimal cosmic than before.
In addition to the simple test function, a test is performed on a more complex function, as shown in fig. 5, the number m of the set population is 30, the number n of the parameters to be optimized is 30, the number of iterations is 1000, and the experimental result is obtained; as shown in fig. 6, the number m of the set population is 30, the number n of the parameters to be optimized is 30, the number of iterations is 1000, and the experimental result is obtained.
It should be understood that the foregoing detailed description of the invention is merely exemplary of the invention and is not intended to limit the invention to the particular forms disclosed. It will be apparent to those skilled in the art that various other modifications and variations can be made in the embodiments of the present invention described above without departing from the spirit and scope of the invention.

Claims (10)

1. An intelligent optimization algorithm of a multi-strategy multi-element universe group is characterized by comprising the following steps:
step S1: setting the number of initial random universes (a plurality of solutions of a problem to be optimized) as m and the number of universe substances (parameters of the problem to be optimized) as n, establishing a universe population matrix U to be optimized in m rows and n columns, and initializing a corresponding universe population matrix E according to the initialized universe population matrix U;
step S2: determining an objective function and a fitness function of a solution, and calculating the fitness f (U) of the universe population U to be optimizedi) And the fitness norm after normalization_f(Ui) And according to the fitness f (U)i) Finding out global optimal cosmic individuals global _ best of the population and historical optimal cosmic best of each cosmic individual in the populationiWhere f is the fitness function, UiThe ith universe in the universe population;
step S3: according to global _ best and bestiUpdating the E universe population matrix according to a certain rule;
step S4: according to the universe individual UiInitializing a new universe individual matrix OiAnd according to global _ best and bestiUpdating the data according to a certain strategy;
step S5: using cosmic object O updated in step S4iUpdating individual E in corresponding E universe populationiAnd record EiIf the iteration times are not changed, judging whether the current iteration E universe individual exceeds a specified number of times and is not updated (if the iteration times exceed the specified number, the current iteration E universe individual is likely to fall into a local optimal solution), if so, screening, covering and updating according to a certain strategy, otherwise, not changing;
step S6: judging whether all the individuals in the current E universe population are completely updated;
if all the updates are completed, the process proceeds to step S7;
if not, returning to the step S4 to continue to carry out individual iteration to update the E universe population;
step S7: calculating and sequencing fitness of all E universe populations updated above for updating universe population U to be optimized;
step S8: for the sorted E universe population, the universe individual E is selected by using a roulette mechanismindexAnd make EindexUpdating the history information of the universe individual and the universe population U to be optimized according to a certain strategyi；
Step S9: judging whether all the individuals in the U universe population are completely updated, if all the individuals in the U universe population are completely updated, entering the next step, and if not, returning to the step S8 to continue the iteration of the universe individuals;
step S10: judging whether a termination condition is reached;
if so, outputting the universe population with the optimal fitness as an optimal value to complete optimization calculation;
and if not, taking the new population matrix as the population matrix to be optimized, and returning to the step S2 to continue iterative optimization.
2. The intelligent multi-policy multivariate cosmic population optimization algorithm according to claim 1, wherein the cosmic population matrix U to be optimized is initialized in step S1 using the Numpy package of Python according to the following codes:
universes＝numpy.zeros((universe_count，universe_dim))
universes[：，i]＝numpy.random.uniform(0，1，m)*(up_list[i]-low_list[i])+low_list[i]
wherein, the universe _ count represents the quantity of the universe population, and the universe _ dim represents the dimension of the problem to be optimized.
3. The intelligent optimization algorithm for multi-policy multi-element universe group according to claim 1, wherein in step S2, the ordered universe group sort _ U-sort (U, F):
wherein, F represents a matrix with m rows and 1 columns of each universe individual fitness, sort (U, F) represents that the fitness matrix is arranged according to the sequence from good to bad, the individuals in the universe group matrix U to be optimized are adjusted according to the sequence from good to bad of the fitness value, the adjusted matrix is output as an ordered universe group matrix sort _ U, and then according to the adjusted ordered universe matrix, global _ best and best are selectedi。
4. The intelligent optimization algorithm for multi-policy multi-element universe group according to claim 1, wherein the universe matrix E is updated in step S3 by the following formula:
Figure FDA0003464301790000021
wherein, c1，r1，d，c2，r2，dAre all in [0,1 ]]Selected uniformly fromAnd (4) counting the machines.
5. The intelligent optimization algorithm for multi-policy multi-element universe group according to claim 1, wherein the universe matrix O is initialized and updated in step S4 by the following formula:
Figure FDA0003464301790000022
Oi，d＝rand(lbd，ubd)，if rd＜pm。
6. the intelligent optimization algorithm for multi-policy multi-universe group according to claim 1, further comprising the following steps in step S5:
universe individual OiUpdating cosmic individuals E by the following formulai：
Figure FDA0003464301790000023
For updated cosmic individuals EiIf the number of times of change exceeds the threshold value of the number of times of no change, 20% of cosmic individuals are selected from the E cosmic population, the cosmic individuals with the best fitness are selected, and the E cosmic individuals to be updated are replacedi(ii) a Otherwise, no change is made.
7. The intelligent optimization algorithm for multi-policy multi-element universe group according to claim 1, further comprising the following steps in step S6:
after updating each universe individual EiIf the individuals which are not updated still exist, returning to the step S4 to update the new E cosmic population individuals; otherwise, the next step is carried out.
8. The intelligent optimization algorithm for multi-policy multi-element universe group according to claim 1, wherein in step S7, the ordered E universe group sort _ E ═ sort (E, F _ E):
wherein, F _ E represents a matrix of m rows and 1 columns of the fitness of each individual in the E cosmic population updated in step S6, and sort (U, F) represents that the fitness matrices are arranged in the order from good to bad, the individuals in the E cosmic population are adjusted in the order from good to bad according to the fitness values, and the adjusted matrix is output as an ordered cosmic population matrix sort _ E.
9. The intelligent optimization algorithm for multi-policy multi-universe group according to claim 1, wherein step S8 further comprises the following steps:
design choice EindexUniverse individual roulette mechanism:
first, the cumulative sum cumsum of ordered E universe population fitness is performed according to the following formulam；
Then, [0, cumsum ] is generatedm]A random number p of the interval;
judging the fitness accumulation and the cumsum of each universe individualiAnd if the index is larger than p, returning the current index which is the E universe individual number selected by the roulette mechanism.
Designing an updating strategy of the universe population U:
Figure FDA0003464301790000024
Figure FDA0003464301790000031
10. the intelligent optimization algorithm for multi-policy multi-universe group according to claim 1, wherein the termination condition in step S10 is any one of the following two conditions:
condition 1: the difference value between the optimal universe individual fitness output by the current iteration and the optimal universe individual fitness output by the last iteration is smaller than a set threshold;
condition 2: the iteration optimization times reach the preset maximum iteration times.
CN202210036297.1A 2022-01-11 2022-01-11 Intelligent optimization algorithm for multi-strategy multi-element universe group Pending CN114444645A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210036297.1A CN114444645A (en) 2022-01-11 2022-01-11 Intelligent optimization algorithm for multi-strategy multi-element universe group

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210036297.1A CN114444645A (en) 2022-01-11 2022-01-11 Intelligent optimization algorithm for multi-strategy multi-element universe group

Publications (1)

Publication Number Publication Date
CN114444645A true CN114444645A (en) 2022-05-06

Family

ID=81368361

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210036297.1A Pending CN114444645A (en) 2022-01-11 2022-01-11 Intelligent optimization algorithm for multi-strategy multi-element universe group

Country Status (1)

Country Link
CN (1) CN114444645A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117439190A (en) * 2023-10-26 2024-01-23 华中科技大学 Water, fire and wind system dispatching method, device, equipment and storage medium

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117439190A (en) * 2023-10-26 2024-01-23 华中科技大学 Water, fire and wind system dispatching method, device, equipment and storage medium
CN117439190B (en) * 2023-10-26 2024-06-11 华中科技大学 Water, fire and wind system dispatching method, device, equipment and storage medium

Similar Documents

Publication Publication Date Title
CN108389211B (en) Image segmentation method based on improved whale optimized fuzzy clustering
WO2022121289A1 (en) Methods and systems for mining minority-class data samples for training neural network
CN109299142A (en) A kind of convolutional neural networks search structure method and system based on evolution algorithm
CN107729999A (en) Consider the deep neural network compression method of matrix correlation
CN109961098B (en) Training data selection method for machine learning
CN110738362A (en) method for constructing prediction model based on improved multivariate cosmic algorithm
CN111105045A (en) Method for constructing prediction model based on improved locust optimization algorithm
CN114724705B (en) Esophageal squamous cell carcinoma survival prediction method based on improved ant lion algorithm and BP neural network
US11914672B2 (en) Method of neural architecture search using continuous action reinforcement learning
CN104504442A (en) Neural network optimization method
WO2022068934A1 (en) Method of neural architecture search using continuous action reinforcement learning
CN113241122A (en) Gene data variable selection and classification method based on fusion of adaptive elastic network and deep neural network
CN115908909A (en) Evolutionary neural architecture searching method and system based on Bayes convolutional neural network
CN110991494A (en) Method for constructing prediction model based on improved moth optimization algorithm
CN111832817A (en) Small world echo state network time sequence prediction method based on MCP penalty function
Zhan et al. Adaptive control of acceleration coefficients for particle swarm optimization based on clustering analysis
CN114444645A (en) Intelligent optimization algorithm for multi-strategy multi-element universe group
Lu et al. Hyper-parameter tuning under a budget constraint
CN111079074A (en) Method for constructing prediction model based on improved sine and cosine algorithm
CN117611974B (en) Image recognition method and system based on searching of multiple group alternative evolutionary neural structures
Elzeki et al. A new hybrid genetic and information gain algorithm for imputing missing values in cancer genes datasets
CN111178416A (en) Parameter adjusting method and device
Cho et al. Genetic evolution processing of data structures for image classification
CN111062524A (en) Scenic spot short-term passenger flow volume prediction method and system based on optimized genetic algorithm
CN109711460A (en) The acquisition methods and device of initial cluster center

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