什么是最快的、MPI 支持的遗传算法 c++图书馆
我想用遗传算法来解决我的特征选择+结构学习问题。由于我们的数据量很大,搜索空间非常大,再加上我们需要做大量的模拟,所以速度非常重要。我发现了几个库,如
什么是最活跃的遗传编程库?< /a>
Java 库用于遗传算法?
以下是我想要的一些功能(按优先级排序):
0) 学术免费
1) 快
2)支持MPI(我们的集群中有MVAPICH2、MPICH2、openMPI)
3) 支持用户定义的基因组和算子
4)综合性(支持精英主义、利基主义、移民等)
5)内置树/图基因组和运算符(可选。如果没有,我将编写自己的。)
6) C++ (可选。我对java或其他语言持开放态度)
7) 支持多目标优化
对于哪一个最适合这些要求有什么建议吗?非常感谢。
I want to use genetic algorithm to solve my feature selection + structure learning problem. Since we have huge amount of data with very large searching space, plus we need to do a lot of simulations, so speed is very important. I found several libraries as in
What is the most active genetic programming library?
Which Java library/libraries for Genetic Algorithms?
But which one is the fastest one?
Here are some features I want (sorted by priority):
0) free for academic
1) fast
2) support MPI (we have MVAPICH2, MPICH2, openMPI in our cluster)
3) support user defined genomes and operators
4) comprehensive (support elitism, niche, migration, etc.)
5) build-in tree/graph genome and operator (optional. If it doesn't, I will write my own.)
6) C++ (Optional. I am open minded to java or other languages)
7) support multiobjective optimization
Any suggestions about which one best fit these requirements? many thanks.
如果你对这篇内容有疑问,欢迎到本站社区发帖提问 参与讨论,获取更多帮助,或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱,如果其他用户或者作者回复了您的评论,将不能在第一时间通知您!
发布评论
评论(1)
您的里程可能会有所不同。
最快的速度取决于您的用例和实施情况。 JGap(java,开源)是一个既定的名称。
基本上,您正在寻找基准?大多数(如果不是全部)已发布的微基准都严重损坏或丢失那些大牌人物。宏观基准通常不太容易被破坏。我与 Drools Planner 竞争(java,开源,没有遗传算法,只有更新的元数据) -一个有趣的启发式方法,例如
模拟退火
)宏观基准测试,其中几个竞争对手使用了遗传算法。所以也许他们的一些决赛论文很有趣。Your mileage may vary.
The fastest one depends on your use case and your implementation. JGap (java, open source) is an established name.
So basically, you're looking for benchmarks? Most (if not all) published micro-benchmarks are seriously broken or missing the big names. Macro-benchmarks are usually less broken. I competed with Drools Planner (java, open source, no genetic algorithms, only more recent meta-heuristics such as
simulated annealing
) in an interesting macro-benchmark where a couple of the competitors used genetic algorithms. So maybe some of their finalist papers are interesting.