mjmckp
mjmckp
There are a few matrix operations that are missing overloads that accept pre-allocated storage for the result. In particular, the `Matrix` instance methods `Transpose`, `Cholesky`, `QR`, and `Evd` suffer from...
This is necessary for PCA, where m >> n and only the first n columns of U are required (if at all). Matlab does this (http://www.mathworks.com.au/help/matlab/ref/svd.html), as does NAG (http://www.nag.com/numeric/MB/manual64_24_1/html/F08/f08kbf.html).
I've created a .NET wrapper for ThunderSvm: https://github.com/mjmckp/ThunderSvmDotNet (also https://www.nuget.org/packages/ThunderSvmDotNet/1.0.0), perhaps you would like to add a link to this in the documentation?
When tuning the SVM parameters, ideally it would be possible to capture the model during training, so that you end up training the model once up to `N` iterations, but...
I am attempting to specify a custom pickler with Thespian, using the following setup code: ``` let registry = CustomPicklerRegistry() registry.RegisterFactory let writer (w: WriteState) (x: MyType) = // code......