[mlpack] mlpack 1.0.9 released

dslate at speakeasy.net dslate at speakeasy.net
Mon Jul 28 19:45:34 EDT 2014


Hi Ryan,

As of Mon Jul 28 18:41:48 CDT 2014, I was able to download mlpack-1.0.9.tar.gz
using the direct link you posted.  However, the mlpack main page at 
http://www.mlpack.org/ also shows this menu entry for getting mlpack 1.0.9:

  download
  mlpack 1.0.9
  released july 28, 2014
  history and news
  gsoc student blog

But when I click on "download" I get mlpack-1.0.8.tar.gz, not mlpack-1.0.9.tar.gz. 
Looks like you need to fix this link. 

Thanks,

-- Dave Slate

On Mon, Jul 28, 2014 at 3:30 PM, Ryan Curtin  wrote:

> Hello there,
>
> Today the last bits of work required for a new mlpack release were
> finished.  You can download mlpack 1.0.9 from
>
>   http://www.mlpack.org/
>
> with a direct URL
>
>   http://www.mlpack.org/files/mlpack-1.0.9.tar.gz
>
> It has been a while since a release, and there have been a lot of
> contributions.  The projects from our five Summer of Code students are
> starting to come to a close, and some of their code has been
> incorporated into this release.  (You can expect another release shortly
> after their projects are done; hopefully early September.)
>
> Here is a list of changes, pulled straight from HISTORY.txt:
>
>  - GMM initialization is now safer and provides a working GMM when
>    constructed with only the dimensionality and number of Gaussians
>    (#314). 
>
>  - Check for division by 0 in Forward-Backward Algorithm in HMMs (#314). 
>
>  - Fix MaxVarianceNewCluster (used when re-initializing clusters for
>    k-means) (#314). 
>
>  - Fixed implementation of Viterbi algorithm in HMM::Predict() (#316). 
>
>  - Significant speedups for dual-tree algorithms using the cover tree
>    (#243, #329) including a faster implementation of FastMKS. 
>
>  - Fix for LRSDP optimizer so that it compiles and can be used (#325). 
>
>  - CF (collaborative filtering) now expects users and items to be
>    zero-indexed, not one-indexed (#324). 
>
>  - CF::GetRecommendations() API change: now requires the number of
>    recommendations as the first parameter.  The number of users in the
>    local neighborhood should be specified with
>    CF::NumUsersForSimilarity(). 
>
>  - Removed incorrect PeriodicHRectBound (#30). 
>
>  - Refactor LRSDP into LRSDP class and standalone function to be
>    optimized (#318). 
>
>  - Fix for centering in kernel PCA (#355). 
>
>  - Added simulated annealing (SA) optimizer, contributed by Zhihao Lou. 
>
>  - HMMs now support initial state probabilities; these can be set in the
>    constructor, trained, or set manually with HMM::Initial() (#315). 
>
>  - Added Nystr=C3=B6m method for kernel matrix approximation by Marcus Ed=
el. 
>
>  - Kernel PCA now supports using Nystr=C3=B6m method for approximation. 
>
>  - Ball trees now work with dual-tree algorithms, via the BallBound<>
>    bound structure (#320); fixed by Yash Vadalia. 
>
>  - The NMF class is now AMF<>, and supports far more types of
>    factorizations, by Sumedh Ghaisas. 
>
>  - A QUIC-SVD implementation has returned, written by Siddharth Agrawal
>    and based on older code from Mudit Gupta. 
>
>  - Added perceptron and decision stump by Udit Saxena (these are weak
>    learners for an eventual AdaBoost class). 
>
>  - Sparse autoencoder added by Siddharth Agrawal. 
>
> If you want a (nearly) full list of bugfixes and changes since 1.0.8,
> you can see the list of resolved tickets:
>
>
> http://www.mlpack.org/trac/query?status=3Dclosed&group=3Dresolution&miles=
tone=3Dmlpack+1.0.9
> (Note: at the moment, Trac is down.  Sorry about that.  It is being
> worked on, but there is no solution quite yet... it should be back up in
> a day or two.)
>
> The benchmarks for 1.0.9 are not done yet, but I will respond to this
> email with a link when they are complete and posted. 
>
> The next release is likely to be 1.1.0; we are anticipating that there
> will be some API reverse compatibility breakage.  However, we will
> provide a guide on any big changes and how code can be adapted.  Many of
> the abstractions we use to organize and understand our algorithms are
> still not completely stable (especially the tree API), and sometimes
> this forces changes... 
>
> Anyway, thanks to everyone for their hard work on this release!  Lots of
> work has gone into this one. 
>
> Ryan
>
> --
> Ryan Curtin    | "This is how Number One works!"
> ryan at ratml.org |   - Number One
> _______________________________________________
> mlpack mailing list
> mlpack at cc.gatech.edu
> https://mailman.cc.gatech.edu/mailman/listinfo/mlpack
>



More information about the mlpack mailing list