ANN: pygene - genetic algorithms package
Hi all, I looked at a few genetic algorithms/genetic programming packages for Python, and found them somewhat convoluted, complicated and counter-intuitive to use. So I've written a genetic algorithms package which I hope will be more approachable to beginners. The first release of pygene is up at: http://www.freenet.org.nz/python/pygene The package includes full api documentation, and an implementation of the travelling salesman problem, plus a couple of simpler cases. -- Cheers aum -- http://mail.python.org/mailman/listinfo/python-list
Re: ANN: pygene - genetic algorithms package
aum wrote: > I looked at a few genetic algorithms/genetic programming packages for > Python, and found them somewhat convoluted, complicated and > counter-intuitive to use. > > So I've written a genetic algorithms package which I hope will be more > approachable to beginners. > > The first release of pygene is up at: > http://www.freenet.org.nz/python/pygene > > The package includes full api documentation, and an implementation of > the travelling salesman problem, plus a couple of simpler cases. I only scanned through the API documentation, but it looks like only genetic algorithms are supported, not full genetic programming. Is this not the case? I've been planning on releasing my stack-based genetic programming system Psi (implemented in Python) at some point in the future, FYI. -- Erik Max Francis && [EMAIL PROTECTED] && http://www.alcyone.com/max/ San Jose, CA, USA && 37 20 N 121 53 W && AIM erikmaxfrancis When in doubt, win the trick. -- Edmund Hoyle -- http://mail.python.org/mailman/listinfo/python-list
Re: ANN: pygene - genetic algorithms package
On Mon, 05 Dec 2005 23:45:30 -0800, Erik Max Francis wrote: > I only scanned through the API documentation, but it looks like only > genetic algorithms are supported, not full genetic programming. Correct. Organisms of a species have a fixed genome. > I've been planning on releasing my stack-based genetic programming > system Psi (implemented in Python) at some point in the future, FYI. Has it got an approachable API? Good doco? Examples understandable by newbies? Hope so. There's a lot of good software that gets ruined because insufficient work has been put in to its usability and comprehensibility. -- Cheers aum -- http://mail.python.org/mailman/listinfo/python-list
Re: ANN: pygene - genetic algorithms package
How is your package different from a nn package? Is this an addon for genetic programming or does it include the standard nn components as well, such as backprop etc? Not being very familiar with genetic programming, forgive me my naive question, I could not immediately find the answer. Thank you, malv -- http://mail.python.org/mailman/listinfo/python-list
Re: ANN: pygene - genetic algorithms package
aum wrote: > Correct. Organisms of a species have a fixed genome. My bad, you had mentioned in the announcement that you had looked at genetic programming systems but didn't claim that pygene was itself a genetic programming system. I misread that; my apologies. >> I've been planning on releasing my stack-based genetic programming >> system Psi (implemented in Python) at some point in the future, FYI. > > Has it got an approachable API? Good doco? Examples understandable by > newbies? Hope so. There's a lot of good software that gets ruined because > insufficient work has been put in to its usability and comprehensibility. That's what I'm working on while polishing it for release. It's also the case that a well-designed class hierarchy will help folks understand the tools that are available to them. "Understandable by newbies" isn't all that a high a priority for me when writing complex software; what's useful is writing software that's easily used by someone familiar with the general concepts, familiar with programming, and familiar with the language that the project is implemented in. You can't teach all things simultaneously; I'm not sure creating a genetic programming (or genetic algorithms) system that's useful to "newbies" (whatever that means) is even a useful goal in and of itself. -- Erik Max Francis && [EMAIL PROTECTED] && http://www.alcyone.com/max/ San Jose, CA, USA && 37 20 N 121 53 W && AIM erikmaxfrancis As far as I'm concerned, being any gender is a drag. -- Patti Smith -- http://mail.python.org/mailman/listinfo/python-list
Re: ANN: pygene - genetic algorithms package
aum wrote: > On Mon, 05 Dec 2005 23:45:30 -0800, Erik Max Francis wrote: >>I only scanned through the API documentation, but it looks like only >>genetic algorithms are supported, not full genetic programming. > > Correct. Organisms of a species have a fixed genome. I've done just enough work in genetic algorithms (and a token amount in genetic programming) to be perplexed by this comment. Are you suggesting that genetic programming is somehow not related to genetic algorithms? My understanding is that (said perhaps somewhat simplistically) genetic programming is an application of genetic algorithms in which the genomes are treated as describing the structure of a program whose execution basically results in the fitness level for that genome. If that's a reasonably accurate statement (or, I suppose, even if it's not), would you please clarify how your "fixed genome" comment relates to any of this? -Peter -- http://mail.python.org/mailman/listinfo/python-list
Re: ANN: pygene - genetic algorithms package
Peter Hansen wrote: > I've done just enough work in genetic algorithms (and a token amount in > genetic programming) to be perplexed by this comment. Are you > suggesting that genetic programming is somehow not related to genetic > algorithms? > > My understanding is that (said perhaps somewhat simplistically) genetic > programming is an application of genetic algorithms in which the genomes > are treated as describing the structure of a program whose execution > basically results in the fitness level for that genome. > > If that's a reasonably accurate statement (or, I suppose, even if it's > not), would you please clarify how your "fixed genome" comment relates > to any of this? You're not replying to me, but I'm the one that elicited that comment. (I was originally asking the question because I misinterpreted the first sentence of his announcement about pygene to mean that pygene was a genetic programming system, but that was never his claim.) A genetic algorithm involves manipulating "programs" which consist of a fixed amount of homogeneous data, for instance, an array of neural network weights, or the coefficients to an equation. Genetic programming involves manipulating general programs, usually as some form of tree. The classic model for genetic programming, from Koza, is where the programs to be manipulated are Lisp s-expressions. pygene implemented a genetic algorithm system, not genetic a programming system, hence his response. It was only my interpretation of his introductory comment that led anyone to believe otherwise. -- Erik Max Francis && [EMAIL PROTECTED] && http://www.alcyone.com/max/ San Jose, CA, USA && 37 20 N 121 53 W && AIM erikmaxfrancis Men live by forgetting -- women live on memories. -- T.S. Eliot -- http://mail.python.org/mailman/listinfo/python-list
Re: ANN: pygene - genetic algorithms package
Erik Max Francis wrote: > You're not replying to me, but I'm the one that elicited that comment. > (I was originally asking the question because I misinterpreted the first > sentence of his announcement about pygene to mean that pygene was a > genetic programming system, but that was never his claim.) > > A genetic algorithm involves manipulating "programs" which consist of a > fixed amount of homogeneous data, for instance, an array of neural > network weights, or the coefficients to an equation. Genetic > programming involves manipulating general programs, usually as some form > of tree. The classic model for genetic programming, from Koza, is where > the programs to be manipulated are Lisp s-expressions. Okay, good, I already knew all that then, except perhaps that key word "fixed". Perhaps I've long been using the wrong label, but I've been doing what I've considered to be "genetic algorithms" and yet working with sometimes variable amounts of sometimes heterogeneous data. I've just considered it to be more sophisticated than the "coefficients in an equation" class of genetic algorithms, but perhaps I've been operating in a gray area between mainstream genetic algorithms and genetic programming. The genomes are certainly not source, nor translatable to source or AST or anything resembling such, in any computer language. Neither, however, could they be described as heterogenous, and for some problems I've been varying the quantity of genetic material in my genomes. Thus my preoccupation with that "fixed". -Peter -- http://mail.python.org/mailman/listinfo/python-list
Re: ANN: pygene - genetic algorithms package
Peter Hansen wrote: > Okay, good, I already knew all that then, except perhaps that key word > "fixed". > > Perhaps I've long been using the wrong label, but I've been doing what > I've considered to be "genetic algorithms" and yet working with > sometimes variable amounts of sometimes heterogeneous data. I've just > considered it to be more sophisticated than the "coefficients in an > equation" class of genetic algorithms, but perhaps I've been operating > in a gray area between mainstream genetic algorithms and genetic > programming. > > The genomes are certainly not source, nor translatable to source or AST > or anything resembling such, in any computer language. Neither, > however, could they be described as heterogenous, and for some problems > I've been varying the quantity of genetic material in my genomes. Thus > my preoccupation with that "fixed". Well, as I'm sure you suspect, there's no "official" definition of either term, and there aren't many rigorous definitions in any case. Coefficients in a function or weights in a neural network are classic examples of a genetic algorithm, but those aren't the only things that would be considered genetic algorithms, although obviously at some point you'd get into some question about what they were. That there are a fixed number of "genes" to be manipulated in general algorithms is also not set in stone (though it is typical). For instance, you might be looking for a function of the form sum_i^n a_i x^i where it might make sense to have n vary in a genetic algorithm system. As a complete distinct example, if you're talking about a virtual machine architecture like Redcode (for Core Wars) with mutation applied (VENUS is an example of this kind of investigation from the early 1990s), then that would probably be considered genetic algorithms, not genetic programming. A general distinction (though not the only one) might be that with genetic algorithms, the manipulation of the data is uniform and completely without reference to its internal makeup. In genetic programming, care has to be taken to preserve program legitimacy. If you went into detail I could probably tell you whether or not what you're doing is "obviously" a genetic algorithm, or "obviously" an instance of genetic programming, or somewhere in between. Without internal structure it's probably more likely closer to genetic algorithms. Not that distinction is that big to begin with; both use very similar concepts in order to get from a random population of programs to a solution that is ideally fit (if possible), including many of the genetic operators (like mutation or crossover) and the way that programs are tested for fitness and how fit or unfit individuals are selected for or against for a new population. That being said, for many problems, it seems that genetic programming has a greater ability to produce solutions to much more complex problems (where often you do not even know the high-level form that they will take -- something you need to decide to put together a genetic algorithm). Recent developments, with stack-based languages like those used by Spector, have allowed the introduction of types naturally into genetic programming, which has a great deal of promise for allowing even more involves solutions to complex problems. -- Erik Max Francis && [EMAIL PROTECTED] && http://www.alcyone.com/max/ San Jose, CA, USA && 37 20 N 121 53 W && AIM erikmaxfrancis I want to know God's thought; the rest are details. -- Albert Einstein -- http://mail.python.org/mailman/listinfo/python-list
Re: ANN: pygene - genetic algorithms package
On Tue, 06 Dec 2005 22:44:39 -0800, Erik Max Francis wrote: > Peter Hansen wrote: > >> Okay, good, I already knew all that then, except perhaps that key word >> "fixed". One thing I should say here is that pygene is a collection of inter-related classes for populations, organisms, gametes and genes. Users are encouraged to subclass off these to address the problems they're trying to solve. For instance, the BaseGene class is completely agnostic about the actual data contained in the gene. Any BaseGene subclass only has to implement the methods: - __add__ - produce phenotype effect of combining a given gene pair - mutate - apply a random mutation to the gene - randomValue - set the gene's value to a legal random value As for the gene's value - as long as the above 3 methods are supplied in a subclass, the gene can hold any object - just as long as one's Organism subclass' '.fitness()' method can understand what the gene class' '.__add__()' method returns. A Gene subclass could even hold an AST for a program to be generated. It wasn't my original intention to support genetic programming, but in theory at least, pygene could be extended into a GP system. Hmm, it's tempting to try a bit of GP, perhaps initially generating code for a tiny language like BrainFuck, then work up to Forth, Factor and beyond. -- Cheers aum -- http://mail.python.org/mailman/listinfo/python-list
Re: ANN: pygene - genetic algorithms package
Erik Max Francis wrote: > If you went into detail I could probably tell you whether or not what > you're doing is "obviously" a genetic algorithm, or "obviously" an > instance of genetic programming, or somewhere in between. Without > internal structure it's probably more likely closer to genetic algorithms. You're certainly correct there. The leap to GP is much farther than my stuff is from GA. >... for many problems, it > seems that genetic programming has a greater ability to produce > solutions to much more complex problems (where often you do not even > know the high-level form that they will take -- something you need to > decide to put together a genetic algorithm). I agree, and I look forward to seeing Psi some time, if just to help me learn more in the area. Thanks again for the comments. -Peter -- http://mail.python.org/mailman/listinfo/python-list
Re: ANN: pygene - genetic algorithms package
Peter Hansen wrote: > You're certainly correct there. The leap to GP is much farther than my > stuff is from GA. I'm curious what sort of heterogeneous data setup you had. > I agree, and I look forward to seeing Psi some time, if just to help me > learn more in the area. Thanks again for the comments. Sure thing. Obviously I'll post an announcement here when it's ready. -- Erik Max Francis && [EMAIL PROTECTED] && http://www.alcyone.com/max/ San Jose, CA, USA && 37 20 N 121 53 W && AIM erikmaxfrancis Heaven and Hell / Is on Earth -- Salt-n-Pepa -- http://mail.python.org/mailman/listinfo/python-list
Re: ANN: pygene - genetic algorithms package
malv wrote: > How is your package different from a nn package? Is this an addon for > genetic programming or does it include the standard nn components as > well, such as backprop etc? > Not being very familiar with genetic programming, forgive me my naive > question, I could not immediately find the answer. No one answered your question, so I will: Neural networks and genetic algorithms/programming are not closely related. A neural network is a model comprising a network of artificial neurons, and signals propagate from inputs, through the neurons of the network (which typically consists of several layers), and finally to the outputs. Algorithms exist to allow neural networks to be trained, so that given a set of inputs and desired outputs, one can iteratively come up with a neural network (of sufficient complexity) that satisfies that criterion. Neural networks can be used for things like pattern recognition and the like. Genetic algorithms/programming is an approach whereby you can start with a population of random programs, apply artificial selection by keeping those that do better at a task than others and genetic modification like mutation and crossover (simulation sexual reproduction) to create diversity, ultimately resulting (hopefully) in programs that are adept at the task. They could be used in conjunction with one another (for instance, a genetic algorithm where the program is represented by an array of neural network weights), but beyond intermixing ideas they really aren't related. -- Erik Max Francis && [EMAIL PROTECTED] && http://www.alcyone.com/max/ San Jose, CA, USA && 37 20 N 121 53 W && AIM erikmaxfrancis I never could have predicted / That I'd feel this way -- India Arie -- http://mail.python.org/mailman/listinfo/python-list
Re: ANN: pygene - genetic algorithms package
Thank you kindly, Erik. malv -- http://mail.python.org/mailman/listinfo/python-list
Re: ANN: pygene - genetic algorithms package
malv wrote: > Thank you kindly, Erik. Sure thing. -- Erik Max Francis && [EMAIL PROTECTED] && http://www.alcyone.com/max/ San Jose, CA, USA && 37 20 N 121 53 W && AIM erikmaxfrancis It is only the poor who are forbidden to beg. -- Anatole France -- http://mail.python.org/mailman/listinfo/python-list
Re: ANN: pygene - genetic algorithms package
Erik Max Francis wrote: > Sure thing. Obviously I'll post an announcement here when it's ready. I'm really happy to see that someone is working on Python based GP implementation :) I'm currently trying to get into GP world (I'm the GP newbie you talked about :P) and, as I'm a Python programmer, I look towards Python based solutions. Unfortunately there are no active Python GP projects (maybe except Pyro, but I'm not sure how GP is implemented there) so I'm forced to play with Java based systems, which isn't what I like. Are you able to give any approximated date of PSI release? Cheers -- eXt -- http://mail.python.org/mailman/listinfo/python-list
Re: ANN: pygene - genetic algorithms package
eXt wrote: > I'm really happy to see that someone is working on Python based GP > implementation :) I'm currently trying to get into GP world (I'm the GP > newbie you talked about :P) and, as I'm a Python programmer, I look > towards Python based solutions. Unfortunately there are no active Python > GP projects (maybe except Pyro, but I'm not sure how GP is implemented > there) so I'm forced to play with Java based systems, which isn't what I > like. > > Are you able to give any approximated date of PSI release? Unfortunately I can't give a precise date. If I have the time, a polished working system with at least the basics should only take a week or so to finish up. Unfortunately, I have a big deadline coming up in my day job, so I'm probably not going to get much time to work on it for the next week or two. Hopefully I'll have a basic system ready by New Year's, but I can't really make any promises. The best way to encourage me to get it done is probably to keep me talking about it :-). -- Erik Max Francis && [EMAIL PROTECTED] && http://www.alcyone.com/max/ San Jose, CA, USA && 37 20 N 121 53 W && AIM erikmaxfrancis You are my martyr / I'm a vestige of a revolution -- Lamya -- http://mail.python.org/mailman/listinfo/python-list
Re: ANN: pygene - genetic algorithms package
Erik Max Francis wrote: > Unfortunately I can't give a precise date. If I have the time, a > polished working system with at least the basics should only take a week > or so to finish up. Unfortunately, I have a big deadline coming up in > my day job, so I'm probably not going to get much time to work on it for > the next week or two. Hopefully I'll have a basic system ready by New > Year's, but I can't really make any promises. Sure, I understand You :). Thanks for answer. > The best way to encourage > me to get it done is probably to keep me talking about it :-). As I did :) -- eXt -- http://mail.python.org/mailman/listinfo/python-list
