TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

What problems have you solved using genetic algorithms/genetic programming?

77 点作者 dlowe超过 14 年前

14 条评论

thornad超过 14 年前
GA solve hard optimization problem, with serious resource constrains, to facilitate breakthrough technology to market:<p>In January 2004 I was contacted by Philips New Display Technologies who were creating the electronics for the first ever commercial e-ink book reader, the Sony Librie, that had only been released in Japan, years before Amazon Kindle and the others hit the market in US an Europe. The Philips engineers had a major problem. Few month before the product was supposed to hit the market, they were still getting ghosting on the screen when changing pages. The problem was the 200 drivers that were creating the electrostatic field. Each of these drivers had a certain voltage that had to be set between zero and 1000 mV or something like this. But if you changed one of them, it would change everything. So optimizing each driver's voltage individually was out of the question. The number of possible combination of values was in billions,and it took about 1 minute for a special camera to evaluate a single combination. The engineers had tried many standard optimization techniques, but nothing would come close. The head engineer contacted me (I was in US at the time, they were in Holland) because I had previously released a Genetic Programming library to the open-source community. He asked if GP/GA's would help and if I can get involved. I did, and for about a month we worked together, me writing and tuning the GA library, on synthetic data, and him integrating it into their system. Then, one weekend they let it run live on the real thing. Next Monday I got these glowing emails from him and their hardware designer, about how nobody could believe the amazing results the GA found. This was it. Later that year the product hit the market. I didn't get payed one cent for it, but I got 'bragging' rights. They said from the begining they were already over budget, so I knew what the deal was before I started working on it. And it's a great story for applications of GAs.
ced超过 14 年前
With regards to GAs, it's good to keep in mind the comments from Russell and Norvig's AI book:<p>"At present, it is not clear whether the appeal of genetic algorithms arises from their performance or from their aesthetically pleasing origins."<p>Furthermore:<p>"Most comparisons of genetic algorithms with other approaches (esp. stochastic hill climbing) have found that the GAs are slower to converge."<p>The discussion at the page below is very enlightening as well.<p><a href="http://books.google.ca/books?id=8jZBksh-bUMC&#38;lpg=PR8&#38;dq=aima&#38;pg=PA155#v=onepage&#38;q=genetic&#38;f=false" rel="nofollow">http://books.google.ca/books?id=8jZBksh-bUMC&#38;lpg=PR8&#38...</a>
评论 #1681916 未加载
评论 #1682305 未加载
评论 #1685521 未加载
评论 #1683034 未加载
endtime超过 14 年前
* Generating rule sets for classifying politicians by party, based on voting data - this was for a class project. IIRC my results were a little better than the standard GrowRule (or whatever it was called) algorithm that most of the class used.<p>* Generating iterated prisoner's dilemma strategies. It's not that this is an intractable problem...but I wanted to see what would happen in a large population of different strategies. This actually resulted in support for some theoretical results I'd read, which was nice. This was for a class paper.<p>* The Mona Lisa polygon thing that Roger Alsing came up with - wrote this just to test a new GA lib I'd written to learn Python.<p>* Optimizing a nano-scale emitter for a new scanning electron microscope - I didn't actually <i>solve</i> this problem, as I had to leave the job where I was working on it before the project finished...but I'd like to think that the project will get somewhere one day.<p>* "Translating" English prose to English verse; that is, trying to increase the metricality of text while minimizing semantic drift. Very, uh, mixed results (I blame Wordnet!). This was also a class project.<p>I feel like I might be forgetting something, but that's at least most of it.
评论 #1682188 未加载
psawaya超过 14 年前
For reference, here's the classic list of 36 known human competitive results produced by GP: <a href="http://www.genetic-programming.com/humancompetitive.html" rel="nofollow">http://www.genetic-programming.com/humancompetitive.html</a>
antirez超过 14 年前
I needed to have a lot of hacking fun: stack based genetic programming provided me with tons of wonderful hacking days a few years ago. See your computer understanding how to create programs like: "DUP * 1 +" actually doing symbolic regression by discovering a program that has the best fitness for the set if inputs is something you'll hardly forget in your life.<p>Btw other than to have fun I also tried to use this approach to create a few hash functions with good distributions.
rouli超过 14 年前
I'm trying to evolve a bot for the current Google's AI challenge using genetic algorithms. The results so far: a mediocre bot that doesn't even win on all maps against the samples bots, and a laptop that is about to die every minute now from exhaustion.
评论 #1682743 未加载
评论 #1682748 未加载
arfrank超过 14 年前
Bill Gross gave a TED talk a few years ago where he discusses how they used genetic algorithms to design a solar collector and a subsequent stirling engine. Theres not much on the details of what went into creating the program, but there were a few interesting pictures of their progress along the way.<p><a href="http://www.ted.com/talks/bill_gross_on_new_energy.html" rel="nofollow">http://www.ted.com/talks/bill_gross_on_new_energy.html</a>
cubicle67超过 14 年前
Solving magic squares<p>I wrote a solver (conventional style) in Ruby that could solve squares in about 30 minutes, and I was pretty pleased with that, so I challenged a friend of mine to beat me.<p>He came back an hour later claiming he had code that could solve it in under 10s. He'd written a genetic algorithm that solved squares on my machine in about 6s <i>and</i> in only about 30 lines of Java (less than my Ruby). I was pretty humbled
motters超过 14 年前
In 2007-9 I developed some software for reading datamatrix patterns. Often these patterns were difficult to read, being indented into scratched surfaces with all kinds of reflectance properties, fuzzy chemically etched markings and so on. I used a GA to fine tune various parameters of the vision algorithms to give the best results on a database of 300 images having known properties. Parameters were things like downsampling resolution, RANSAC parameters, amount of erosion and dilation, low pass filtering radius, and a few others. Running the optimisation over several days this produced results which were about 20% better than naive values on a test set of images unseen during the optimisation phase.<p>This system was completely written from scratch, and I didn't use any other libraries. I'm not opposed to using such things, provided that they give a reliable result, but you have to be careful about license compatibility and code portability issues.
dangrossman超过 14 年前
I used some GP to find optimal ranges for sensors meant to detect malware on computers. The sensors were things like process and thread counts, memory usage, disk activity, etc... but there were around 40 total sensors and very large ranges of data seen on computers running normally and computers with some kind of malware running in the background. The detection software built with the GP results worked pretty well. It was just a project I did for a reverse engineering class in college.
terra_t超过 14 年前
GA and GP are powerful methods for machine learning, but other methods are often a better use of programmer and computer time, particularly the Support Vector Machine.
jfager超过 14 年前
Answering a slightly different question: if you're considering trying to solve a problem with a GA, you might first try simulated annealing. For many problems, it will probably converge to a decent answer more quickly, is quite a bit easier to code, and is less resource intensive.
abecedarius超过 14 年前
Made <a href="http://djb.deviantart.com/gallery/" rel="nofollow">http://djb.deviantart.com/gallery/</a> using my eye for the fitness function. Karl Sims did it better originally, but I had some fun with it.
c00p3r超过 14 年前
<i>My first job as a professional programmer (1995) was writing a genetic-algorithm based automated trading system for S&#38;P500 futures. The application was written in Visual Basic 3</i> - thanks! You made my day!