Evolutionary Algorithms
From ResearchID.org
|
Introduction (?)
Evolutionary algorithms are a computational tool that has been used both for studying the abilities and limitations of evolution in given situations and for harnessing those abilities.
Evolution in its most general form requires the following:
- A population of organisms
- Reproduction of the organisms
- Mutation of some feature of the organisms
- A "fitness function" based on that feature
- Selective culling of the population based on their fitness
In the world around us, the organisms are living creatures or viruses, the critical feature of the organism is its DNA, and the fitness function is the organism's ability to survive and reproduce in its environment. In an evolutionary algorithm, the organisms might be field-programmable gate arrays, the critical feature might be the FPGA's logic states and the fitness function might be the FPGA's success at a particular task.
Evolutionary algorithms have been successfully employed in computer science and engineering, among other fields, to produce highly efficient programs and designs for a vast array of tasks. Their ability to successfully search through large solution spaces varies heavily depending on the shape of the fitness function - for example, an evolutionary algorithm would fare poorly at a task that had only one "right" solution, but would often do much better than a human at a task that had a large quantity of partially "right" solutions.
Making an evolutionary algorithm that accurately reflects real-world evolution is something of a challenge. Setting up a fitness function that is too tightly-defined will limit the effectiveness of evolutionary techniques such as co-option. Keeping the fitness function static will lead to stagnation. Limiting the simulation to only one "species" will eliminate all chance of seeing behaviours such as sexual dimorphism that are believed to only be advantageous in the presence of multiple mutually-antagonistic species. Currently the evolutionary algorithm that is most "realistic" in this sense (whilst not being computationally impractical) is probably Avida, which is based on the old Core Wars multiplayer hacker game.
Type (?)
Level (?)
Graduate, Doctorate, Postdoctorate
Definition (?)
Explore the limitations of real-world evolution by analogy with computer programs possessing the necessary conditions for evolution to occur.
Objectives (?)
- Build up a small library of evolutionary algorithms against which theoretically-derived claims such as Dembski's Displacement Theorem and Conservation of Information can be tested
- Experimentally determine the limitations of evolutionary algorithms of particular varieties in particular circumstances; compare with real world to determine areas where evolution may be insufficient
Questions (?)
- How can evolutionary algorithms be made more accurate in their reflection of real-world evolution?
- Under what circumstances do evolutionary algorithms succeed and fail?
- What restrictions would this place on the initial conditions for life?
- It has been claimed that macroevolution and microevolution are fundamentally different phenomena rather than simply the same phenomenon viewed at different orders of magnitude. Does the behaviour of genetic algorithms support or undermine this claim?
Keywords (?)
--- ! Begin keywords instructions ! ---
- Provide a simple list of words directly related to Evolutionary Algorithms. (3-15 words)
--- ! End keywords instructions ! ---
Monographs (?)
--- ! Begin monographs instructions ! ---
- 1-10 of the most relevant published texts. This could be books, articles, or any hard copy resource. Cite in this format:
Title of Resource, Author Name, ISBN (number)
- Be sure to include the characters 'ISBN' before entering the actual number, that way we can compile a site-wide bibliography by doing a search for 'ISBN'.
- Including the ISBN number also gives a hyperlink to sites where they can get more info on the book and purchase it.
- If Peer Reviewed writings, please designate type of peer review:
- (p.r.) indicates that a work is peer reviewed by a mainstream journal
- (p.r.I.D.) indicates that a work is peer reviewed by an ID journal
--- ! End monographs instructions ! ---
eResources (?)
--- ! Begin eResources instructions ! ---
- Evolving 3D Morphology and Behavior by Competition An evolutionary simulation of populations of artificial creatures with evolved morphology and behaviour. The fitness function is based upon creature's ability to stay in contact with a goal object relative to other members of the population.
Criticism (?)
--- ! Begin criticism instructions ! ---
- The most useful critiques of Evolutionary Algorithms, included here as a simple title with a hyperlink or citation.
- This section is not for the inclusion of a critical abstract, only links or citations.
- (1-10 links or citations)
--- ! End criticism instructions ! ---
Status (?)
Current
Extend (?)
--- ! Begin extend instructions ! ---
- Other questions that follow from this application that would be worthy of investigation.
- Entries in the 'Extend' section could have high potential for becoming new research applications.
- (700 characters or less)
--- ! End extend instructions ! ---
Related (?)
--- ! Begin1 related instructions ! ---
- Links to other research applications at ResearchID.org that are related to this this article about Evolutionary Algorithms. (1-25 links)
- Don't forget that these could be links to biography, glossary, or FreeStyle pages.
- Feel free to add links here to related research that does not have a page on ResearchID.org yet!
--- ! End related instructions ! ---
Notes (?)
--- ! Begin notes instructions ! ---
- How to use "wikified" citations Step-by-step:
- Where you want the citation to appear in the article, type in: {{ref|descriptionname}}, replacing 'descriptionname' with a unique name of the footnote.
- In the "Notes" section, type in: # {{note|descriptionname}}
- All of the notes under this section should be surrounded by these "div style" tags in order to make the text smaller:
- <div style="font-size: 90%">...Notes and citations here...</div>
- Be sure to remove the 'nowiki' tags all the way through, if they ended up in your wiki code.
- Use the citation and notation style of choice for the field, or the style that you would prefer use. Consistency within the article is the goal. If you want some ideas on style, check out Wikipedia's citation style examples.
--- ! End notes instructions ! ---
Field(s) (?)
-- ! Begin Field(s) instructions ! ---
- Here at ResearchID.org, academic fields are arranged and organized into categories.
- List at least one category related to the study of Evolutionary Algorithms. You can select them from the categories page or make a new category!
- For example, if I want this page to be listed in the "Biology" and "Astrobiology" category, I would type in:
[[Category:Biology]] [[Category:Astrobiology]]
- Be sure to remove the 'nowiki' tags and the three spaces before the opening tag. They are only added here in order to ensure they appear for demonstration purposes on the template page as regular text, and not as functioning category links.
- To make a new category, type in an unused category name in the above format.
- Once a category is properly added, a simple list of the categories will appear at the bottom of the research article.
- Learn more at the Help:Categories page.
--- ! End Field(s) instructions ! ---
