<p dir="ltr">Congratulations all on 2.0.0 release!<br>
Do we have a page on who and where all is mlpack used ?</p>
<br><div class="gmail_quote"><div dir="ltr">On Thu, Dec 24, 2015, 08:20 Ryan Curtin <<a href="mailto:ryan@ratml.org">ryan@ratml.org</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Hello there,<br>
<br>
This has been a long time coming...<br>
<br>
Last night I tagged mlpack-2.0.0 and uploaded it to the mlpack website.<br>
You can get it here:<br>
<br>
<a href="http://www.mlpack.org/files/mlpack-2.0.0.tar.gz" rel="noreferrer" target="_blank">http://www.mlpack.org/files/mlpack-2.0.0.tar.gz</a><br>
<br>
There has been a significant amount of refactoring and hard work by lots<br>
of people since the last release in January, and the changelog is fairly<br>
long, so I'll put what I think are the most exciting bits below:<br>
<br>
* Parallelization: the DET (density estimation trees) code is now<br>
parallelized with OpenMP. As time goes on, parallelization will be<br>
added to other algorithms, but note that you can also use Armadillo<br>
with OpenBLAS, which will parallelize all the linear algebra calls.<br>
<br>
* Model saving and loading: where appropriate, all of the command-line<br>
programs now support loading and saving models. So you can train,<br>
say, a logistic regression model, and save it for later use. This is<br>
also possible with techniques like all-k-nearest-neighbor search,<br>
which allow you to save the tree built on the points. Model<br>
serialization support is also available from C++, too, of course.<br>
<br>
* Significant refactoring: most machine learning algorithms now follow<br>
the same API, and documentation has been improved.<br>
<br>
* Tree-based algorithms now support multiple types of trees in a far<br>
easier manner.<br>
<br>
* The k-means code now supports five different algorithms, many of them<br>
far faster than the original implementation.<br>
<br>
* Add streaming decision trees (Hoeffding trees) for fast classifiers<br>
on huge datasets. This supports both categorical and numeric<br>
features.<br>
<br>
* No more dependence on libxml2; boost::serialization is used instead.<br>
<br>
* Armadillo minimum version bump to 4.100.0.<br>
<br>
* All mlpack programs are now prefixed with 'mlpack_', so for instance<br>
'allknn' is now 'mlpack_allknn'.<br>
<br>
Also exciting, in my opinion, is the community that has grown around<br>
mlpack. Here are some neat and interesting statistics:<br>
<br>
* mlpack has almost 40 contributors<br>
<br>
* mlpack has now been downloaded at least 35k+ times (my logs<br>
undercount)<br>
<br>
* mlpack has been used in at least 40 academic papers (also a lowball<br>
estimate)---and this number is increasing faster and faster<br>
<br>
* the mlpack codebase now contains about 60k source lines of code<br>
(SLOC)<br>
<br>
So I have to say, I'm very happy that we have built tools that people<br>
are finding useful! I hope that this trend continues. :)<br>
<br>
For the full changelog in mlpack-2.0.0, see<br>
<a href="http://www.mlpack.org/history.html" rel="noreferrer" target="_blank">http://www.mlpack.org/history.html</a>. Over the next few days/weeks,<br>
updated mlpack packages will be pushed to the package repositories of<br>
various distributions.<br>
<br>
Lastly, some notes about the future. Upcoming releases will follow the<br>
versioning guidelines now present in UPDATING.txt (semantic versioning):<br>
<a href="https://github.com/mlpack/mlpack/blob/master/UPDATING.txt" rel="noreferrer" target="_blank">https://github.com/mlpack/mlpack/blob/master/UPDATING.txt</a><br>
<br>
Future goals include a flexible framework for artificial neural networks<br>
(prototype code can currently be found in the master branch in<br>
src/mlpack/methods/ann), generic bindings to other languages such as<br>
Python, Java, MATLAB, and others, parallelization support for more<br>
algorithms via OpenMP, a new implementation of random forests, and<br>
dimensionality reduction or manifold learning techniques.<br>
<br>
I'm also hopeful that we can have a much more frequent release cycle,<br>
more like once a month or more, following the versioning guidelines I<br>
mentioned earlier.<br>
<br>
So, I hope that you find this release useful! Please feel free to<br>
report any bugs as Github issues to <a href="https://github.com/mlpack/mlpack" rel="noreferrer" target="_blank">https://github.com/mlpack/mlpack</a> or<br>
to this mailing list, or to the #mlpack channel in freenode.<br>
<br>
--<br>
Ryan Curtin | "Good Lord - I've heard about this - cat juggling!<br>
<a href="mailto:ryan@ratml.org" target="_blank">ryan@ratml.org</a> | Stop! Stop! Stop it!" - Navin R. Johnson<br>
_______________________________________________<br>
mlpack mailing list<br>
<a href="mailto:mlpack@cc.gatech.edu" target="_blank">mlpack@cc.gatech.edu</a><br>
<a href="https://mailman.cc.gatech.edu/mailman/listinfo/mlpack" rel="noreferrer" target="_blank">https://mailman.cc.gatech.edu/mailman/listinfo/mlpack</a><br>
</blockquote></div>