[shark] 01/79: updated documentation to reflect 3.0 release

Ghislain Vaillant ghisvail-guest at moszumanska.debian.org
Thu Nov 26 15:39:05 UTC 2015


This is an automated email from the git hooks/post-receive script.

ghisvail-guest pushed a commit to branch master
in repository shark.

commit d576bf7045f8c9c4f914f87558e949dc8d7a68aa
Author: Oswin <oswin.krause at di.ku.dk>
Date:   Mon Oct 26 09:06:35 2015 +0100

    updated documentation to reflect 3.0 release
---
 doc/sphinx_pages/index.rst                         |  4 -
 doc/sphinx_pages/rest_sources/about_shark/news.rst |  8 +-
 .../rest_sources/downloads/downloads.rst           | 10 ++-
 .../rest_sources/getting_started/installation.rst  |  6 +-
 .../tutorials/concepts/data/datasets.tut           |  4 +-
 .../tutorials/concepts/misc/statistics.rst         | 96 +++++++++++++---------
 .../tutorials/concepts/misc/statistics.tut         |  2 +-
 .../tutorials/first_steps/shark_layout.rst         |  9 +-
 .../rest_sources/tutorials/tutorials.rst           |  1 -
 9 files changed, 76 insertions(+), 64 deletions(-)

diff --git a/doc/sphinx_pages/index.rst b/doc/sphinx_pages/index.rst
index 7e2b19b..d79164a 100644
--- a/doc/sphinx_pages/index.rst
+++ b/doc/sphinx_pages/index.rst
@@ -114,10 +114,6 @@
 Summary
 =======
 
-.. note::
-
-     This is Shark 3.0 beta. See the :doc:`news <rest_sources/about_shark/news>` for more information.
-
 **SHARK is a fast, modular, feature-rich open-source C++ machine learning library**.
 It provides methods for linear and nonlinear optimization, kernel-based learning
 algorithms, neural networks, and various other machine learning techniques (see the
diff --git a/doc/sphinx_pages/rest_sources/about_shark/news.rst b/doc/sphinx_pages/rest_sources/about_shark/news.rst
index b16f5fe..9c7469d 100644
--- a/doc/sphinx_pages/rest_sources/about_shark/news.rst
+++ b/doc/sphinx_pages/rest_sources/about_shark/news.rst
@@ -1,8 +1,14 @@
 News
 ====
 
+
+Shark 3.0.0 Released
+^^^^^^^^^^^^^^^^^^^^^^^^
+
+We are happy to announce the release of Shark 3.0.0
+
 Shark moves to GitHub
-^^^^^^^^^^^^^^^
+^^^^^^^^^^^^^^^^^^^^^^^^
 
 Today, Shark moved to GitHub, please update your repositories, see the downloads page for more details.
 
diff --git a/doc/sphinx_pages/rest_sources/downloads/downloads.rst b/doc/sphinx_pages/rest_sources/downloads/downloads.rst
index c877d4b..c235e13 100644
--- a/doc/sphinx_pages/rest_sources/downloads/downloads.rst
+++ b/doc/sphinx_pages/rest_sources/downloads/downloads.rst
@@ -40,12 +40,14 @@ Get the latest snapshot from our svn repository!
 
 
 
-.. Shark sources
-   -------------
+Shark sources
+-------------
 
-   Please download the following package if you want to build Shark yourself:
+   We have two source packages available:
 
-   `Shark source code <https://nisys.dyndns.biz/shark/job/Shark_Source_Package/lastSuccessfulBuild/artifact/libshark-3.0.0-src.tar.bz2>`_
+   `Shark-3.0.0.zip <https://github.com/Shark-ML/Shark/archive/v3.0.0.zip>`_
+   
+   `Shark-3.0.0.tar.gz <https://github.com/Shark-ML/Shark/archive/v3.0.0.tar.gz>`_
 
    See the :doc:`installation guide <../getting_started/installation>`
    for details on how to compile and install the library.
diff --git a/doc/sphinx_pages/rest_sources/getting_started/installation.rst b/doc/sphinx_pages/rest_sources/getting_started/installation.rst
index 76eb436..609180d 100644
--- a/doc/sphinx_pages/rest_sources/getting_started/installation.rst
+++ b/doc/sphinx_pages/rest_sources/getting_started/installation.rst
@@ -15,16 +15,14 @@ Packages
 We are supporting packages for the following platforms:
 
 * Arch Linux
-	- current development version downloadable from AUR as packages ``shark-ml-atlas-svn`` and ``shark-ml-svn``
+	- current development version downloadable from AUR as packages ``shark-ml-atlas-git`` and ``shark-ml-git``
 	  for a shark version with and without ATLAS.
 * More to come!
 
 Installation
 ---------------------------------------------
 
-To install Shark, get the sources::
-
-	git clone https://github.com/Shark-ML/Shark/
+To install Shark, get the sources from our * :doc:`Downloads page <downloads>`
 	
 Then build the library::
 
diff --git a/doc/sphinx_pages/rest_sources/tutorials/concepts/data/datasets.tut b/doc/sphinx_pages/rest_sources/tutorials/concepts/data/datasets.tut
index b341ee9..d1804ae 100644
--- a/doc/sphinx_pages/rest_sources/tutorials/concepts/data/datasets.tut
+++ b/doc/sphinx_pages/rest_sources/tutorials/concepts/data/datasets.tut
@@ -258,7 +258,7 @@ batches since it needs to find the batch the element is located in.
 .. warning::
 	Element-wise random access to a ``Data`` a object is a linear time
 	operation! It is not to be confused with constant-time element
-	access in arrays (and ``std::vector``). Thus aside from only very
+	access in arrays. Thus aside from only very
 	small data sets or performance uncritical code you should never use
 	element-wise random-access to a data container.
 
@@ -377,7 +377,7 @@ It is applied to the data set by calling: ::
 ..sharkcode<Data/Datasets.tpp,transform-4>
 
 .. note::
-	Never ever forget the definition of the ``result_type``!
+	Never never forget the definition of the ``result_type``!
 	It is needed by ``transform`` to be smart, i.e., to deduce
 	the corresponding batch type.
 	If you happen to get nasty template error messages with
diff --git a/doc/sphinx_pages/rest_sources/tutorials/concepts/misc/statistics.rst b/doc/sphinx_pages/rest_sources/tutorials/concepts/misc/statistics.rst
index d812c66..37db2a6 100644
--- a/doc/sphinx_pages/rest_sources/tutorials/concepts/misc/statistics.rst
+++ b/doc/sphinx_pages/rest_sources/tutorials/concepts/misc/statistics.rst
@@ -3,63 +3,79 @@ Iterative Calculation of Statistics
 
 The Shark machine learning library includes a component for
 iteratively calculating simple descriptive statistics of a
-sequence of values. The class :doxy:`Statistics` is a thin
-wrapper around the boost::accumulators component. This tutorial
-illustrates its usage.
+sequence of points for experimental evaluation. The class :doxy:`ResultTable`
+includes a simple data aggregation tool that for a set of experiments
+with different parameters  aggregates results over a set of trials. It 
+supports missing values to reflect failed trials as well.
+The class :doxy:`Statistics` takes these results to cpmpute a set of statistics.
+The class offers the possibility to label the dimensions of the points and statistics
+to automatically generate human readable output, for example in a csv table.
 
-We need the following header files: ::
+For this simple application, we are going to generate some points from
+a gaussian distribution and then mark some points as missing.
+For this experiment, we need the following header files: ::
 
 
 	#include <shark/Statistics/Statistics.h>
 	#include <shark/Rng/GlobalRng.h>
 	
 
-We start out by creating an object of class :doxy:`Statistics`: ::
+We start out by creating an object of class :doxy:`ResultTable`.
+We give the table a name and also label the inputs as to generate 
+a more readable output later on::
 
 
-		Statistics stats;
+		statistics::ResultTable<double> table(2,"VarianceOfGaussian");//set a name for the results
+		table.setDimensionName(0,"input1");
+		table.setDimensionName(1,"input2");
 	
 
-Now we feed numbers into this object. For demonstration purposes we
-sample 100,000 i.i.d. standard normally distributed values from the
-random number generator. ::
 
 
-		// Sample 10000 standard normally distributed random numbers
-		// and update statistics for these numbers iteratively.
-		for (std::size_t i = 0; i < 100000; i++)
-			stats( Rng::gauss() );
+Now we feed numbers into this object. For demonstration purposes we
+sample 10,000 i.i.d. standard normally distributed values with varying
+variance. To simulate a failed experiment, we make a coin toss for variable two
+and mark this input as missing. Finally, we insert the values into the table::
+
+
+		// Fill the table with randomly generated numbers for different variances and mean and also add missing values
+		for(std::size_t k = 1; k != 10; ++k){
+			double var= 10.0*k;
+			for (std::size_t i = 0; i < 10000; i++){
+				double value1=Rng::gauss(0,var);
+				double value2=Rng::gauss(0,var);
+				if(Rng::coinToss() == 1)
+					value2=statistics::missingValue();
+				table.update(var,value1,value2 );
+			}
+		}
 	
 
-We can output a summary to the console: ::
+Next, we generate a :doxy:`Statistics` object and add the statistics, here
+we use Mean, Variance and Percentage of Missing values::
 
 
-		// Output results to the console.
-		std::cout << stats << std::endl;
+		statistics::Statistics<double> stats(&table);
+		stats.addStatistic(statistics::Mean());//adds a statistic "Mean" for each variable
+		stats.addStatistic("Variance", statistics::Variance());//explicit name
+		stats.addStatistic("Missing", statistics::FractionMissing());
 	
 
-The results looks similar to: ::
-
-	Sample size: 100000
-	Min: -4.09568
-	Max: 4.42802
-	Mean: -0.000584566
-	Variance: 0.992282
-	Median: 0.00121767
-	Lower Quantile: -0.673034
-	Upper Quantile: 0.670621
-
-Alternatively it is possible to access the individual values. The
-round bracket operator is used for this purpose: ::
-
-
-		std::cout << 
-			stats( Statistics::NumSamples() ) << " " <<
-			stats( Statistics::Min() ) << " " <<
-			stats( Statistics::Max() ) << " " <<
-			stats( Statistics::Mean() ) << " " << 
-			stats( Statistics::Variance() ) << " " <<
-			stats( Statistics::Median() ) << " " <<
-			stats( Statistics::LowerQuartile() ) << " " <<
-			stats( Statistics::UpperQuartile() ) << std::endl;
+We can output a summary as csv file to the console: ::
+
+
+		printCSV(stats);
 	
+
+The results looks similar to::
+
+	# VarianceOfGausian Mean-input1 Mean-input2 Variance-input1 Variance-input2 Missing-input1 Missing-input2
+	10 0.00500042 -0.073452 9.77016 10.1016 0 0.5061
+	20 0.0359687 -0.0400334 20.1318 20.2767 0 0.5038
+	30 0.0216264 -0.120275 30.3096 29.0293 0 0.5044
+	40 -0.0301033 0.0995221 40.3523 40.4839 0 0.4961
+	50 0.00692523 0.118349 48.9781 50.5156 0 0.4936
+	60 -0.0133728 -0.0109795 57.4287 59.8386 0 0.4903
+	70 -0.190326 0.0259554 67.0553 70.0034 0 0.4987
+	80 -0.0198076 -0.0493343 83.1629 78.0985 0 0.4917
+	90 -0.103546 -0.263991 92.152 89.3462 0 0.4992
diff --git a/doc/sphinx_pages/rest_sources/tutorials/concepts/misc/statistics.tut b/doc/sphinx_pages/rest_sources/tutorials/concepts/misc/statistics.tut
index 499b274..2e817a6 100644
--- a/doc/sphinx_pages/rest_sources/tutorials/concepts/misc/statistics.tut
+++ b/doc/sphinx_pages/rest_sources/tutorials/concepts/misc/statistics.tut
@@ -35,7 +35,7 @@ and mark this input as missing. Finally, we insert the values into the table::
 Next, we generate a :doxy:`Statistics` object and add the statistics, here
 we use Mean, Variance and Percentage of Missing values::
 
-..sharkcode<Statistics/Statistics.tpp,statisticse>
+..sharkcode<Statistics/Statistics.tpp,statistics>
 
 We can output a summary as csv file to the console: ::
 
diff --git a/doc/sphinx_pages/rest_sources/tutorials/first_steps/shark_layout.rst b/doc/sphinx_pages/rest_sources/tutorials/first_steps/shark_layout.rst
index bc88145..8ccbdc7 100644
--- a/doc/sphinx_pages/rest_sources/tutorials/first_steps/shark_layout.rst
+++ b/doc/sphinx_pages/rest_sources/tutorials/first_steps/shark_layout.rst
@@ -26,9 +26,8 @@ choices:
   and also sets up a :doxy:`Data` class especially suited for
   machine learning tasks: subsets (e.g., for cross-validation) are
   lazy copies of the original set.
-* The folders ``Fuzzy/``, ``Network/``, ``Rng/``, ``Statistics/`` all
-  implement specialized functionality pertaining to Fuzzy Logic, HTTP
-  protocols (for RESTful APIs), random number generation, and various
+* The folders `Rng/`` and ``Statistics/``
+  implement specialized functionality for random number generation, and various
   statistical tests or distributions, respectively.
 
 * Currently, the only algorithms implemented in the folder ``Unsupervised/``
@@ -38,7 +37,3 @@ choices:
   ``"Model"``-``"ObjectiveFunction"``-``"Optimizer"`` trias. This can be seen as roughly
   corresponding to the three remaining folders ``Models/``, ``ObjectiveFunctions/``,
   and ``Algorithms/``
-
-.. todo::
-  This mindmap will be extended in scope and provided as an interactive,
-  browsable applet in the official release of Shark.
diff --git a/doc/sphinx_pages/rest_sources/tutorials/tutorials.rst b/doc/sphinx_pages/rest_sources/tutorials/tutorials.rst
index 5d5f5cf..04c23d1 100644
--- a/doc/sphinx_pages/rest_sources/tutorials/tutorials.rst
+++ b/doc/sphinx_pages/rest_sources/tutorials/tutorials.rst
@@ -163,7 +163,6 @@ Sharks comes with its own solver for Quadratic Programs:
 .. _label_for_linalg_tutorials:
 
 We give an introduction to Shark's usage of the
-`Boost uBLAS <http://www.boost.org/doc/libs/release/libs/numeric>`_
 library for "all things linear algebra":
 
 * :doc:`concepts/lin_alg/vector_matrix`

-- 
Alioth's /usr/local/bin/git-commit-notice on /srv/git.debian.org/git/debian-science/packages/shark.git



More information about the debian-science-commits mailing list