New doc from Joerg

git-svn-id: http://root.cern.ch/svn/root/trunk@37549 27541ba8-7e3a-0410-8455-c3a389f83636

tmva/doc/v528/index.html

-<br> 
-<hr> 
-<a name="tmva"></a> 
-<h3>TMVA</h3>
+    <br>
+    <hr>
+    <a name="tmva"></a>
+    <h3>TMVA</h3>
 
-    TMVA version 4.1.0 is included in this root release.
+    TMVA version 4.1.0 is included in this root release. The most
+    important new feature is the support for simulataneous
+    classification for multiple classes for a number of multi-variate
+    methods. A lot of effort went into consolidation of the software,
+    i.e. method performance and robustness, and framework
+    stability. The changes with respect to ROOT 5.27 / TMVA 4.0.7 are
+    in detail:
 
     <h4>Framework</h4>
 
     <ul>
+      <li><b><em>Multi-class support.</em></b> The support of multiple
+      output classes (i.e., more than a single background and signal
+      class) has been enabled for these methods: MLP (NN), BDTG,
+      FDA.<br>
+
+        <div style="margin-left:20; font-size: small">The multiclass
+        functionality can be enabled with the Factory option
+        <code>"AnalysisType=multiclass"</code>. Training data is
+        specified with an additional classname, e.g. via
+        <code>factory->AddTree(tree,"classname");</code>. After the
+        training a genetic algorithm is invoked to determine the best
+        cuts for selecting a specific class, based on the figure of
+        merit: purity*efficiency. TMVA comes with two examples in
+        <code>$ROOTSYS/tmva/test</code>: <code>TMVAMulticlass.C</code>
+        and <code>TMVAMulticlassApplication.C</code></div>
+
+      </li>
+
       <li> <b><em>New TMVA event vector building.</em></b> The code
       for splitting the input data into training and test samples for
       all classes and the mixing of those samples to one training and
       one test sample has been rewritten completely. The new code is
-      more performant and has a clearer structure. This should fix
-      several bugs which have been reported by
+      more performant and has a clearer structure. This fixes several
+      bugs which have been reported by some users of TMVA.</li>
+
+      <li><b><em>Code and performance test framework</em></b>: A unit
+      test framework for daily software and method performance
+      validation has been implemented.
+      </li>
+
+    </ul>
+
+    <h4>Methods</h4>
+
+    <ul>
+
+      <li><b><em>BDT Automatic parameter optimisation for building the
+      tree architecture</em></b>: The optimisation procedure uses the
+      performance of the trained classifier on the "test sample" for
+      finding the set of optimal parameters. Two different methods to
+      traverse the parameter space are available (scanning, genetic
+      algorithm). Currently parameter optimization is implemented only
+      for these three parameters that influence the tree architectur:
+      the maximum depth of a tree, <code>MaxDepth</code>, the minimum
+      number of events in each node, <code>NodeMinEvents</code>, and
+      the number of tress, <code>NTrees</code>.
+
+        <div style="margin-left:20; font-size: small">Optimization can
+        is invoked by calling
+        <code>factory->OptimizeAllMethods();</code> prior to the call
+        <code>factory->TrainAllMethods();</code>.</div>
+
+        Automated and configurable parameter optimization is soon to
+        be enabled for all methods (for those parameters where
+        optimization is applicable).
+      </li>
+
+      <li>
+        <b><em>BDT node splitting</em></b>: While Decision Trees
+        typically have only univariate splits, in TMVA one can now
+        also opt for multivariate splits that use a "Fisher
+        Discriminant" (option: UseFisherCuts), built from all
+        observables that show correlations larger than some threshold
+        (MinLinCorrForFisher).  The training will then test at each
+        split a cut on this fisher discriminant in addition to all
+        univariate cuts on the variables (or only on those variables
+        that have not been used in the Fisher discriminant, option
+        UseExcusiveVars). No obvious improvement betwen very simple
+        decision trees after boosting has been observed so far, but
+        only a limited number of studies has been performed concerning
+        potiential benenfit of these simple multivariate splits.
+      </li>
 
     </ul>
 
+
+    <h4>Bug fixes</h4>
+
+    <ul>
+      <li>A problem in the BDTG has been fix, leading to a much
+      improved regression performance.</li>
+
+      <li>With the new test framework and the coverity checks of ROOT
+      a number of bugs were discovered and fixed. The mainly concerned
+      memory leaks, and did not really affect performance.</li>
+
+    </ul>
+
+
+
+
+
+
+