[mlpack-git] [mlpack] [Proposal]Enhance the class SparseAutoencoder and SoftmaxRegression (#454)

[stereomatchingkiss]

>I didn't completely understand this idea, could you elaborate more please?
Ok, I mean provide a default constructor which takes the name of the files

    SoftmaxRegression(std::string const &fileName);

This way the users can reuse the training results in their codes easier(without training).

    SoftmaxRegression<> sm("smoke_params_of_sm"); 
    sm.Predict(data, result);

>Replace it with what? Do you mean moving it to src/mlpack/methods/ann/ or something?
Replace it with the original files under the folder method/sparse_autoencoder
Since backward compatibility is not an issue, I will place them int the folder "method/sparse_autoencoder"

>I think maybe instead of default constructors, simple constructors that don't take data can be used.
Thanks, will reference from it

>Why not just use the existing data::Load() function that works with boost::serialization once Serialize() is implemented?
Thanks:), I did not noticed this solution



---
Reply to this email directly or view it on GitHub:
https://github.com/mlpack/mlpack/issues/454#issuecomment-143475336
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/mlpack-git/attachments/20150926/b46db267/attachment.html>