[mlpack-git] [mlpack] improve speed of SparseAutoencoder and make it more flexible (#451)

[stereomatchingkiss]

Finally, I can run the Train function without crash, but the training results is correct or good enough or not is another question, need more test.

Sample codes : 

    const arma::mat trainData = arma::randu<arma::mat>(2,4);
    const size_t visibleSize = trainData.n_rows;
    const size_t hiddenSize = trainData.n_rows / 2;    

    using AnnSae = ann::LogisticSparseAutoencoder<>;
    size_t const batch_size = trainData.n_cols;
    size_t const epoch = 100;
    AnnSae  sae(visibleSize, hiddenSize, batch_size);
    sae.Train(trainData, epoch);



---
Reply to this email directly or view it on GitHub:
https://github.com/mlpack/mlpack/pull/451#issuecomment-166157389
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/mlpack-git/attachments/20151220/dec59023/attachment.html>