Pre-Grant Publication Number: 20080016013
Please help the USPTO examine the application by evaluating the relevance of the publicly submitted prior art to the patent application.
Peer-to-Patent forwards the Top 10 most relevant prior art submissions and their annotations to the United States Patent and Trademark Office.
Review this prior art and click on the thumbs up (or down) to indicate whether this submission should be forwarded to the USPTO.
If you login then you can add an annotation by typing in the box at the bottom of the screen to comment on the relevance of the prior art to the claims of the patent application.
Review this prior art and click on the thumbs up (or down) to indicate whether this submission should be forwarded to the USPTO.
If you login then you can add an annotation by typing in the box at the bottom of the screen to comment on the relevance of the prior art to the claims of the patent application.
Prior Art Detail
Summary / Description
| Summary / Description | In claim 1, the inventors claim a particular decomposition of a MOEA that mirrors the generally understood structure of the algorithm. Object-oriented software systems typically use the same or very similar decompositions. The decomposition described in this paper is identical except that it groups the fitness evaluator and the dominance filter into one component (see figure 1). The random number generator is included by not illustrated. The paper also allows archiving of fitness evaluations according to user specifications. |
Basic Information
| Type of Prior Art | Print Publication |
| Publication Title * | Computational Intelligence and Multimedia Applications, 1999. ICCIMA apos;99. Proceedings. Third Int |
| Author | Kumar, R.; Kumar, N.V.; Nagrath, I.J. |
| ISBN | |
| Page Range | 96-100 |
| Medium | Other printed publication |
| Publication Date * | January 1, 1999 |
| URL | http://ieeexplore.ieee.org/Xplo... |
Notes / To Do
| Notes | |
Excerpt
Excerpt Please see figure 1. In addition:
Chromosomes can be represented as binary strings or real-valued numbers. They can be of fixed or variable length. They can be built from any C++ data types - bit strings, arrays, lists or tress - using the types built-in to the library. Random number generator routines [6] are provided for initialisation and generating random points at various stages of evolution. Both generational and steady-state populations are supported. Elitism is optional for generational type of GAs so that a fraction of the best-fit individuals are copied into the mating pool. A bit-tag is attached to each individual of the population which indicates its non-dominating or dominating position on the Pareto front. Selection and replacement strategies are provided. Search operators include a variety of crossover and mutation operators which also include the constraints for variable length representations.
A user can code his objective functions and opt for the genetic strategy. The population can be ranked on Pareto-front and fitness score calculated. For constraint optimisation, the toolkit design imposes constraints on variables, and supports penalising the objectives and zooming-out the Pareto-front in a region of interest. A stopping criterion can be defined - to run the GA for some fixed number of iterations, to terminate the optimisation when some fraction of the population has become non-dominated, or assessing the convergence of Pareto-front using rankhistograms.
The run-time statistics of population evolutions and objective function evaluations can be recorded by setting which statistics should be recorded and how often they should be. Most commonly used statistics are maximum, minimum, mean, standard deviation, movement of Pareto-front, age-distribution of the nondominated solutions and the rank-histograms. GA parameters can be configured from file, code and/or user-interface. User can also fine-tune the
parameters while viewing the run-time performance and analysing the generational statistics. |
Relevance
Claims
1
A system for implementing a multi objective evolutionary algorithm (MOEA) on a programmable hardware device, the system comprising:
a random number generator, configured to generate a sequence of pseudo random numbers;
a population generator; configured to generate a population of solutions based on the output from the random number generator;
a crossover/mutation module, configured to adapt the population of solutions to generate an adapted population of solutions;
a fitness evaluator configured to evaluate each member comprising the population of solutions and the adapted population of solutions, wherein the fitness evaluator is implemented on the programmable hardware device;
a dominance filter configured to select a subset of members from the population of solutions and the adapted population of solutions and generate a filtered population of solutions; and
an archive configured to store populations of solutions.
Relevance
The paper describes a typical system with nearly the same generalized components that, together, are used for the same purpose. The key difference is that the paper uses a single component, "Objective Evaluation Pareto-ranking, & Fitness score/scaling," instead of the combination of "a fitness evaluator" and "a dominance filter." The same functions are identified, but not explicitly separated in the paper. The system in the paper allows archiving of fitness evaluations and presumably the corresponding solutions (the common practice) according to user specifications.
The paper describes a typical system with nearly the same generalized components that, together, are used for the same purpose. The key difference is that the paper uses a single component, "Objective Evaluation Pareto-ranking, & Fitness score/scaling," instead of the combination of "a fitness evaluator" and "a dominance filter." The same functions are identified, but not explicitly separated in the paper. The system in the paper allows archiving of fitness evaluations and presumably the corresponding solutions (the common practice) according to user specifications.
Claim Chart
All
12
A method for implementing a multi objective evolutionary algorithm (MOEA) on a programmable hardware device, the method comprising:
generating a sequence of pseudo random numbers and generating a population of solutions based on the sequence of pseudo random numbers;
adapting the population of solutions to generate an adapted population of solutions;
evaluating each member of the population of solutions and the adapted population of solutions, wherein the evaluation is performed using a fitness evaluator implemented on the programmable hardware device;
selecting a subset of members from the population of solutions and the adapted population of solutions to generate a filtered population of solutions, wherein the steps of evaluating and selecting are iteratively performed until one or more termination criteria are reached; and
archiving the populations of solutions.
Relevance
The method involves invoking the functions of the components in an order consistent with the commonly understood structure of the genetic algorithm. Since these functions correspond to the steps of the genetic algorithm, this is not novel. In the paper, this in captured in the arrangement of the components in figure 1.
The method involves invoking the functions of the components in an order consistent with the commonly understood structure of the genetic algorithm. Since these functions correspond to the steps of the genetic algorithm, this is not novel. In the paper, this in captured in the arrangement of the components in figure 1.
Claim Chart
All
