[mlpack-git] [mlpack] implement command line programs of softmaxRegression (#466)

Ryan Curtin notifications at github.com
Sun Nov 8 21:01:11 EST 2015


> why arma::vec& but not arma::Row?

The Armadillo storage format is column-major, so inside mlpack C++ code, a column (either arma::mat::col() or arma::vec, arma::Col<T>, etc.) refers to a data point.

> To work with SGD, I have to implement three more functions

Nah, you shouldn't have to do anything at all.  You can use the SoftmaxRegression class with the L-BFGS optimizer like this:

`SoftmaxRegression<L_BFGS>`

and with the SGD optimizer like this:

`SoftmaxRegression<SGD>`

So you shouldn't need to modify any of the existing code.  Take a look at `src/mlpack/methods/nca/nca_main.cpp` and how the --optimizer option is implemented there; it should be like that, pretty straightforward.  I hope that clarifies things a bit.

> We could rely on the huge libraries developed by NVDIA and other open source community. Like  mshadow(https://github.com/dmlc/mshadow) and Thrust(http://docs.nvidia.com/cuda/thrust/).

> mshadow provide high level api for matrix operation like armadillo, Thrust provide stl like general purpose algorithms.

Excellent, I will take a look into these when I have some time.  Thanks for the links!

---
Reply to this email directly or view it on GitHub:
https://github.com/mlpack/mlpack/pull/466#issuecomment-154902110
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/mlpack-git/attachments/20151108/e5a74bdb/attachment.html>


More information about the mlpack-git mailing list