F# 1.9.9.9 和 lapack
谁能帮我举例说明如何使用 lapack 和最新的 f# 吗？或者暂时不推荐？
我搜索后只找到 http://fdatamining.blogspot.com/ 但无法获取 FSharp.PowerPack.Math.Providers.dll 已编译。
编辑： @Yin，谢谢你的博客。从 1.9.7.8 源代码编译 FSharp.PowerPack.Math.Providers.dll 时，多次使用 Microsoft.FSharp.Compatibility.permutation 并且无法找到。在您的博客中据说只是明确定义它：
type permutation = int –> int.
或者我将所有 Microsoft.FSharp.Compatibility.permutation 替换为 Microsoft.FSharp.Math.Permutation 并对其进行编译。还没有测试过。
您能否向我们展示 svd bug 的测试用例？
谢谢
Can anyone help me with example on using lapack with latest f#? Or it is not recommended for now?
I searched and only find http://fdatamining.blogspot.com/ but was not able to get FSharp.PowerPack.Math.Providers.dll compiled.
EDIT:
@Yin, Thanks for your blog. On compiling FSharp.PowerPack.Math.Providers.dll from 1.9.7.8 source, Microsoft.FSharp.Compatibility.permutation is used several times and cannot be found. In your blog it is said just to define it explicitly:
type permutation = int –> int.
Alternatively I replace all Microsoft.FSharp.Compatibility.permutation with Microsoft.FSharp.Math.Permutation and got it compiled. Have not test it yet.
Could you please show us the test case for svd bug?
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）
我写了这个博客。你用VS 2008吗？你安装了PowerPack吗？到目前为止你做了什么？你得到了什么编译错误？
F# 团队不建议使用数学提供程序。然而，这是我能在网上找到的最好的 F# 版本（不完整且 alpha 测试的 Math.Net 需要时间才能成熟）。我还可以告诉你，包装器有问题，例如 SVD 包装器中有一个错误。但这个包装器很小，您可以在 F# 新版本中自行维护它，并根据需要添加新功能（我们这里有明显的许可证问题！）。我的计划是使用和修改这个包装器直到它稳定并重写一个新的包装器以解决许可证问题。或者，我们可以说服MS或MSR继续开发。
@ahala，svd bug
搜索
成员 this.dgesvd_((a:matrix)) =
在 lapack_service_netlib.fs 中，
在 m×n(m>n) 矩阵上执行 SVD 时，该函数存在内存访问/段错误错误。
为了解决这个问题，我们需要将arg_jobu和arg_jobvt更改为“S”。参考 lapack 文档：http://linux.die.net/man/l/dgesvd
I wrote this blog. Do you use VS 2008? Have you installed PowerPack? What have you done so far? What compile errors did you get?
Using the math providers is NOT recommended by the F# team. However, this is the best I can find online for F# (The incomplete and alpha-test Math.Net needs time to mature). I can also tell you that the wrapper is buggy, e.g. there's a bug in the SVD wrapper. But this wrapper is small, you can maintain it yourself along the F# new releases and add new functions as you need (we have obvious license problem here!). My plan is to use and modify this wrapper until it is stable and rewrite a new one to clear license problem. Or, we can persuade MS or MSR to continue the development.
@ahala, the svd bug
search
member this.dgesvd_((a:matrix)) =
in lapack_service_netlib.fs
This function has a memory-access/segment-fault bug when doing SVD on a m-by-n(m>n) matrix.
To solve the problem, we need to change arg_jobu and arg_jobvt to 'S'. ref lapack doc here: http://linux.die.net/man/l/dgesvd