Thread View: comp.ai.genetic
2 messages
2 total messages
Started by Eric Siegel
Fri, 22 Jan 1993 21:04
Reproduction Methods
Author: Eric Siegel
Date: Fri, 22 Jan 1993 21:04
Date: Fri, 22 Jan 1993 21:04
52 lines
2803 bytes
2803 bytes
pauld@cs.washington.edu (Paul Barton-Davis) says: > One really interesting feature that I have observed from my own work > with a tierra-like system is the effect of a reproductive burden on > the speed of evolution. ... > So, in normal GA-land, you simply use the "hand of god" > approach: you reach in, and *you* do the reproduction, eliminating any > selection pressure on this feature. ... > But wait ! Perhaps, just perhaps, this reproductive feature is just > what you want. Its not very difficult to make a tierra-like system > evolve its own form of crossover. All kinds of other wierd and > wonderful reproductive mechanisms evolve too. It seems possible to me > that when evolution is viewed as explorations of a vast state space, > particular reproductive methods might be incapable of ever searching > some particular areas of that space. Leaving the system free to devise > its own reproductive strategies opens up the possibility of exploring > areas of the state space that "programmer/user-selected" methods might > not. Solution: A GA organism does not only represent a "vector" in the solution space of the domain problem, but also has (seperately) encoded a "vector" in _reproduction method space_ (e.g. Weights of crosss-over, mutation; Number of cross-over points, etc. A big research area?) This way, the evolutionary method which converges most quickly for the domain at hand will be converged to! And this way, you don't have to strain so hard to decided how much of biology to simulate. Yeah, I was just thinking about evolution a month ago, and said to my friend over lunch, "you ever realize that the manner by which evolution takes place evolves?" To go a step further, for systems that use GAs on-line (such as schedulers?), use GA technology to evolve the evolutionary method _off-line_. --------- This comparison between tierra-like systems and GAs reminds me: In the former, an effective upper-limit on time spent running an organism's program is implicit in the organism's environment. In the latter, the Halting Problem prevents us from using all-out models of computation. But I haven't heard these things mentioned explicitely... Haven't they been addressed? --------- Has there been any work using evolutionary systems to teach a computer English? (I'm in Natural Language Processing.) References appreciated.
Re: Reproduction Methods
Author: Paul Barton-Davis
Date: Fri, 22 Jan 1993 22:15
Date: Fri, 22 Jan 1993 22:15
57 lines
3302 bytes
3302 bytes
In article <C19x7D.DK3@cs.columbia.edu> evs@cs.columbia.edu (Eric Siegel) writes: >pauld@cs.washington.edu (Paul Barton-Davis) says: > >> One really interesting feature that I have observed from my own work >> with a tierra-like system is the effect of a reproductive burden on >> the speed of evolution. ... > >> So, in normal GA-land, you simply use the "hand of god" >> approach: you reach in, and *you* do the reproduction, eliminating any >> selection pressure on this feature. ... > >> But wait ! Perhaps, just perhaps, this reproductive feature is just >> what you want. Its not very difficult to make a tierra-like system >> evolve its own form of crossover. All kinds of other wierd and >> wonderful reproductive mechanisms evolve too. It seems possible to me >> that when evolution is viewed as explorations of a vast state space, >> particular reproductive methods might be incapable of ever searching >> some particular areas of that space. Leaving the system free to devise >> its own reproductive strategies opens up the possibility of exploring >> areas of the state space that "programmer/user-selected" methods might >> not. > >Solution: A GA organism does not only represent a "vector" in the >solution space of the domain problem, but also has (seperately) >encoded a "vector" in _reproduction method space_ (e.g. Weights >of crosss-over, mutation; Number of cross-over points, etc. A big >research area?) Well, not really. This still leaves the programmer to devise an appropriate set of reproductive methods. The beauty of tierra is that (1) no fitness function for reproductive fitness is needed, because it is implicit in the ranking of the genome within the population and (2) any reproductive method anyone could ever dream up can potentially evolve within the system, as well as few that no-one ever would. >Yeah, I was just thinking about evolution a month ago, and said to my >friend over lunch, "you ever realize that the manner by which evolution >takes place evolves?" My precise point. Programmers trying to second-guess what makes it tick seems like an unnecessary constraint. >This comparison between tierra-like systems and GAs reminds me: In the >former, an effective upper-limit on time spent running an organism's >program is implicit in the organism's environment. Want to explain what you mean ? I don't see such an upper-limit, other than the one imposed by the realities of the host system (ie. real world physics). -- paul -- hybrid rather than pure; compromising rather than clean; | Militant Agnostic distorted rather than straightforward; ambiguous rather than| I Don't Know articulated; both-and rather than either-or; the difficult | and You Don't unity of inclusion rather than the easy unity of exclusion. | Know Either
Thread Navigation
This is a paginated view of messages in the thread with full content displayed inline.
Messages are displayed in chronological order, with the original post highlighted in green.
Use pagination controls to navigate through all messages in large threads.
Back to All Threads