Systems and methods for auto-adaptive control over converged results for multi-dimensional optimization
원문보기
IPC분류정보
국가/구분
United States(US) Patent
등록
국제특허분류(IPC7판)
G06N-003/12
G06F-017/30
출원번호
US-0194424
(2011-07-29)
등록번호
US-8862627
(2014-10-14)
발명자
/ 주소
Ferringer, Matthew Phillip
Thompson, Timothy Guy
출원인 / 주소
The Aerospace Corporation
대리인 / 주소
Sutherland Asbill & Brennan LLP
인용정보
피인용 횟수 :
0인용 특허 :
10
초록▼
Systems and methods may include identifying an input population of parent epsilon chromosome data structures; combining genes of each selected pair of parent epsilon chromosome data structures according to at least one evolutionary operator to generate a plurality of child epsilon chromosome data st
Systems and methods may include identifying an input population of parent epsilon chromosome data structures; combining genes of each selected pair of parent epsilon chromosome data structures according to at least one evolutionary operator to generate a plurality of child epsilon chromosome data structures, each child epsilon chromosome data structure providing one or more genes each having a respective candidate epsilon value representing a respective step size or spacing for the respective problem objective; and evaluating each of the plurality of child epsilon chromosome data structures according to one or more epsilon objective functions to generate respective epsilon objective function values for each child epsilon chromosome data structure, where each epsilon objective function is associated with a respective goal associated with at least one a priori criterion, where each respective epsilon objective function value indicates an extent to which each respective goal can be achieved.
대표청구항▼
1. A method comprising: identifying an input population of parent epsilon chromosome data structures, wherein each parent epsilon chromosome data structure provides genes each having a respective candidate epsilon value, each candidate epsilon value representing a respective step size or spacing ass
1. A method comprising: identifying an input population of parent epsilon chromosome data structures, wherein each parent epsilon chromosome data structure provides genes each having a respective candidate epsilon value, each candidate epsilon value representing a respective step size or spacing associated with a respective problem objective of a plurality of problem objectives;selecting one or more pairs of parent epsilon chromosome data structures from the input population of parent epsilon chromosome data structures;combining genes of each selected pair of parent epsilon chromosome data structures according to at least one evolutionary operator to generate a plurality of child epsilon chromosome data structures, each child epsilon chromosome data structure providing one or more genes each having a respective candidate epsilon value representing a respective step size or spacing for the respective problem objective; andevaluating each of the plurality of child epsilon chromosome data structures according to one or more epsilon objective functions to generate respective epsilon objective function values for each child epsilon chromosome data structure, wherein each epsilon objective function is associated with a respective goal associated with at least one a priori criterion defined using at least a respective subset of the plurality of problem objectives, wherein each respective epsilon objective function value indicates an extent to which each respective goal can be achieved, wherein the prior steps are performed by one or more computerswherein the identifying, selecting, combining, and evaluating steps form an epsilon optimization process, and further comprising:selecting an epsilon chromosome data structure from one of the evaluated chromosome data structures, wherein candidate epsilon values from the selected epsilon chromosome data structure form an epsilon vector utilized in performing a problem optimization process, wherein the problem optimization process seeks to identify a set of epsilon non-dominated solutions for each respective subset of the plurality of problem objectives, wherein a size of the set of solutions for each subset is based at least in part on the epsilon vector. 2. The method of claim 1, wherein the respective a priori criterion is associated with a number of desired solutions within a respective problem search space defined by the respective subset of the plurality of problem objectives. 3. The method of claim 2, wherein the respective problem search space defined by the respective subset of the plurality of problem objectives is a total search space defined by an entirety of the problem objectives. 4. The method of claim 2, wherein the evaluating comprises sorting an archive of possible problem solutions using the respective step size or spacing of the respective candidate epsilon value of the respective child epsilon chromosome structure and evaluating the extent to which each respective goal of the number of desired solutions within the respective problem search space can be achieved. 5. The method of claim 1, wherein the identified set of solutions is stored in an archive. 6. The method of claim 1, wherein the epsilon optimization process is triggered following a termination of a first run of the problem optimization process, wherein upon completion of the epsilon optimization process, the epsilon vector is utilized in performing a second run of the problem optimization process. 7. The method of claim 1, wherein the epsilon vector is a second epsilon vector, wherein the first run of the optimization process utilizes a first epsilon vector, wherein respective epsilon values of the second epsilon vector vary within one or more predefined ranges from epsilon values of the first epsilon vector. 8. The method of claim 1, wherein epsilon values in the epsilon vector are utilized in performing the second run of the second optimization process. 9. The method of claim 1, wherein the respective a priori criterion is associated with a minimum spacing requirement of solutions within a respective problem search space defined by the respective subset of the plurality of problem objectives. 10. The method of claim 1, wherein a number of subsets of the plurality of problem objectives being utilized for a problem optimization process is equal to a total number of the one or more epsilon objective functions. 11. The method of claim 1, wherein each problem objective is associated with a minimization or a maximization of a respective dimension. 12. The method of claim 1, wherein the at least one evolutionary operator includes one or both of a cross-over operator or a mutation operator. 13. The method of claim 1, wherein at least a portion of the identified input population of parent epsilon chromosome data structures is randomly generated. 14. A system comprising: at least one memory that stores computer-executable instructions; andat least one processor configured to access the at least one memory, wherein the at least one processor is configured to execute the computer-executable instructions to:identify an input population of parent epsilon chromosome data structures, wherein each parent epsilon chromosome data structure provides genes each having a respective candidate epsilon value, each candidate epsilon value representing a respective step size or spacing associated with a respective problem objective of a plurality of problem objectives;select one or more pairs of parent epsilon chromosome data structures from the input population of parent epsilon chromosome data structures;combine genes of each selected pair of parent epsilon chromosome data structures according to at least one evolutionary operator to generate a plurality of child epsilon chromosome data structures, each child epsilon chromosome data structure providing one or more genes each having a respective candidate epsilon value representing a respective step size or spacing for the respective problem objective; andevaluate each of the plurality of child epsilon chromosome data structures according to one or more epsilon objective functions to generate respective epsilon objective function values for each child epsilon chromosome data structure, wherein each epsilon objective function is associated with a respective goal associated with at least one a priori criterion defined using at least a respective subset of the plurality of problem objectives, wherein each respective epsilon objective function value indicates an extent to which each respective goal can be achievedwherein the identifying, selecting, combining, andevaluating from an epsilon optimization process, and wherein the at least one processor is further configured to execute the computer-executable instructions to:select an epsilon chromosome data structure from one of the evaluated chromosome data structures, wherein candidate epsilon values from the selected epsilon chromosome data structure form an epsilon vector utilized in performing a problem optimization process, wherein the problem optimization process seeks to identify a set of epsilon non-dominated solutions for each respective subset of the plurality of problem objectives, wherein a size of the set of solutions for each subset is based at least in part on the epsilon vector. 15. The system of claim 14, wherein the respective a priori criterion is associated with a number of desired solutions within a respective problem search space defined by the respective subset of the plurality of problem objectives. 16. The system of claim 15, wherein the evaluation is performed by sorting an archive of possible problem solutions using the respective step size or spacing of the respective candidate epsilon value of the respective child epsilon chromosome structure and evaluating the extent to which each respective goal of the number of desired solutions within the respective problem search space can be achieved. 17. The system of claim 14, wherein the respective a priori criterion is associated with a minimum spacing requirement of solutions within a respective problem search space defined by the respective subset of the plurality of problem objectives. 18. The system of claim 14, wherein each problem objective is associated with a minimization or a maximization of a respective dimension.
연구과제 타임라인
LOADING...
LOADING...
LOADING...
LOADING...
LOADING...
이 특허에 인용된 특허 (10)
Tolson Michael (Mill Valley CA), Image processing using genetic mutation of neural network parameters.
Ferringer, Matthew Phillip; Clifton, Ronald Scott; Thompson, Timothy Guy, Systems and methods for a core management system for parallel processing of an evolutionary algorithm.
Ferringer, Matthew Phillip; Clifton, Ronald Scott; Thompson, Timothy Guy, Systems and methods for parallel processing optimization for an evolutionary algorithm.
Ferringer, Matthew Phillip; Thompson, Timothy Guy; Clifton, Ronald Scott; DiPrinzio, Marc David, Systems and methods for supporting restricted search in high-dimensional spaces.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.