genetic algorithm completely rewritten

ACKNOWLEDGEMENTS

@@ -19,11 +19,11 @@ Tim DiLauro
 Robert Ferguson
 Andrew Hankinson
 Thomas Karsten
-J�rgen Killian
+Jürgen Killian
 Jonathan Koch
 David Kolanus
 Uma Kompella
-Jean-Fran�ois Leon
+Jean-François Leon
 Andreas Leuschner
 David Lewis
 Moritz Muras

CHANGES

@@ -1,6 +1,23 @@
 Changes made between Gamera File Releases
 =========================================
 
+Version 3.4.0, Nov 15, 2012
+----------------------------
+
+ - GA optimization module rewritten from scratch
+     o evolutionary optimization now for selection and weighting implemented
+     o replacement of used library (GALib was replaced by EO)
+     o optimization process is now highly configurable
+     o fitness evaluation is now parallelizable
+
+ - knncoremodule is now splitted in header and implemenation file
+
+ - feature normalization for noninteractive classifier now works
+
+ - feature selection supported as a better alternative to feature weighting
+
+ - Point-types can be added
+
  - new plugins brink_threshold and diagonal_projections (feature)
    (thanks to Andrew Hankinson)
 

MANIFEST.in

@@ -5,6 +5,7 @@ recursive-include doc/html *.css
 recursive-include gamera/pixmaps *.png *.py *.bmp
 recursive-include gamera/test *.tiff
 recursive-include src *.cpp *.c *.h makefile.* *.hpp *.txt ANNOUNCE CHANGES INSTALL LICENSE README TODO KNOWNBUG Y2KINFO
+recursive-include src/eodev *
 recursive-include examples *.py README *.tiff *.xml
 recursive-include include *.hpp *.hxx *.h README
 recursive-include win32 *.py

TODO

@@ -7,10 +7,6 @@ ToDo list for Gamera
    algorithms. Maybe some parts of Leptonica could be useful?
  - Add multilevel thresholding.
  - Support FFT on Gamera images by wrapping FFT from VIGRA
- - Feature selection via genetic algorithms (GA) as an alternative
-   to the currently implemented GA based feature weighting.
-   Investigate why the GA based feature weighting in Gamera produces
-   so poor results.
  - A new menu function in the interactive classifier dialog for
    highlighting outliers, i.e. possible manual misclassifications.
    Can be based on MNN editing.

doc/src/classify.txt

@@ -50,17 +50,17 @@ particular classifier type.  The currently implemented classifiers
 are all *k* - nearest-neighbor, but we plan to add other classifiers
 as needed.
 
- - kNNInteractive_
+ - kNNInteractive
      Interactive *k* nearest neighbor classifier.
 
- - kNNNonInteractive_
+ - kNNNonInteractive
      Noninteractive *k* nearest neighbor classifier.  The weights of
      the dimensions can be optimised using a genetic algorithm.
 
-     For a user-friendly way to perform GA optimization, consider the Biollante_
-     tool in the Gamera GUI.
+     For a user-friendly way to perform GA optimization, consider the module_
+     documentation.
 
-.. _Biollante: gui.html#classifier-optimization-biollante
+.. _module: ga_optimization.html
 
 Image properties related to classification
 ------------------------------------------
@@ -339,13 +339,8 @@ classifier API.  It adds some methods of its own.
   ``area``.  This normalization may change the classifications that
   the classifier makes, however.
 
-- The weights of the features can be optimized using a genetic
-  algorithm.  (See start_optimizing_).
-
-  For a user-friendly way to perform GA optimization, consider the Biollante_
-  tool in the Gamera GUI.
-
-.. _Biollante: gui.html#classifier-optimization-biollante
+- The selections and weights of the features can be optimized using a genetic
+  algorithm (See here_).
 
 - The training data can be serialized to a classifier-specific binary file
   format.  This format saves and loads much faster than the Gamera XML
@@ -356,6 +351,8 @@ classifier API.  It adds some methods of its own.
    file, since it is portable across platforms and to future versions of
    Gamera.  The binary format is not guaranteed to be portable.
 
+.. _here: ga_optimization.html
+
 Feature management
 ''''''''''''''''''
 
@@ -401,6 +398,11 @@ the following settings are plain properties of all kNN classifier classes:
 
 .. docstring:: gamera.knn kNNInteractive change_feature_set
 
+.. docstring:: gamera.knn _kNNBase get_selections_by_features get_selections_by_feature
+.. docstring:: gamera.knn _kNNBase get_weights_by_features get_weights_by_feature
+.. docstring:: gamera.knn _kNNBase set_selections_by_features set_selections_by_feature
+.. docstring:: gamera.knn _kNNBase set_weights_by_features set_weights_by_feature
+
 .. docstring:: gamera.knn _kNNBase save_settings load_settings
 
 Serialization
@@ -419,18 +421,6 @@ Evaluation
 
 .. docstring:: gamera.knn kNNInteractive noninteractive_copy
 
-``kNNNonInteractive``
-'''''''''''''''''''''
-
-Optimization
-````````````
-
-These methods pertain to optimizing the feature weights of the classifier
-using a genetic algorithm (GA). For information about genetic algorithms,
-see [Holland1975]_. The GA implementation in Gamera is based on [GAlib]_.
-
-.. docstring:: gamera.knn kNNNonInteractive start_optimizing stop_optimizing add_optimization_callback remove_optimization_callback
-
 Improving kNN Classifiers using Editing
 '''''''''''''''''''''''''''''''''''''''
 
@@ -493,15 +483,3 @@ Integrating your own editing algorithm
 If you have written your own editing algorithm and want to make it available in
 the Classifier GUI, refer to the documentation of the class 
 **gamera.knn_editing.AlgoRegistry**.  
-
-
-References
-----------
-
-.. [Holland1975] Holland, J. H. 1975. *Adaptation in natural and
-   artifical systems.*  University of Michigan Press, Ann Arbor.
-
-.. [GAlib] Wall, M. 1999. `GAlib. A C++ Library of Genetic Algorithm
-    Components.`__
-
-.. __:  http://lancet.mit.edu/ga/

doc/src/directory_heirarchy.txt

@@ -56,8 +56,9 @@ The Gamera source tree looks something like this:
 | src/                | Source code (that compiles directly to object                  |
 |                     | files.)                                                        |
 |                     +---------------+------------------------------------------------+
-|                     | ga/           | GALib_: Matthew's Genetic Algorithm Library    |
-|                     |               | Used to optimize the kNN classifier.           |
+|                     | eodev/        | EOdev_: Evolving Objects (EO): an              |
+|                     |               | Evolutionary Computation Framework             |
+|                     |               | used to optimize the kNN classifier.           |
 |                     +---------------+------------------------------------------------+
 |                     | graph/        | A graph library.  API should be considered     |
 |                     |               | unstable.                                      |
@@ -86,6 +87,6 @@ The Gamera source tree looks something like this:
 .. _docutils: http://docutils.sf.net/
 .. _Writing Gamera Plugins: writing_plugins.html
 .. _VIGRA Computer Vision Library: http://kogs-www.informatik.uni-hamburg.de/~koethe/vigra/
-.. _GALib: http://lancet.mit.edu/ga/
+.. _EOdev: http://eodev.sourceforge.net/
 .. _libtiff: http://www.libtiff.org/
 .. _script-writing documentation: script.html