Change log
1.3.0.92 (Nov 12 2016)
- Added multi-threaded evaluation of equations in the outer genetic algorithm loop. Added option '--maxThreads' that controls the number of concurrent execution threads used.
1.2.4.83 (Mar 23 2014)
- Deprecated the "avg" (average) operator - it is redundant because the basic multiplication, sum, and constant operators achieve the same mathematical expression.
- Fixed a bug in parsing tabular text files where the first column contains numerical values instead of textual row headers.
1.2.3.74 (Apr 18 2012)
- Fixed a bug in fitting a linear equation with a constant term.
1.2.2.70 (Mar 10 2012)
- Fixed a bug in calculating standard deviation with option 'doScaling'.
1.2.1.61 (Nov 22 2011)
- Added 64-bit versions for Windows and Linux.
- Minor code cleanup.
1.2.0.47 (Wed 14 Jan 2009)
- Changed program output to include the correlation index R2 and Friedman's lack of fit (LOF) measure for the fit to the calibration data.
- Added option 'FriedmansSmooth' for the smoothing parameter value used in computing LOF.
- Added options 'maxNDescriptors' and 'minNDescriptors' that (optionally) control the number of descriptors allowed in the generated equations. The default value, maxNDescriptors = 0, imposes no limit.
1.1.1.42 (Wed 10 Dec 2008)
- Fixed a bug that may have caused an infinite loop in cross-validation.
1.1.0.35 (Fri Jul 18 2008)
- Added options for performing y-scrambling.
- Added option 'linear' that forces the generation of linear equations.
- Added option for scaling input data to zero mean and unit standard deviation.
- Added calculation of relative leverage values. See the Features page for more information on leverage.
- Activity file can also be read in a prediction run in order to compute residuals.
- Fixed a bug in reading Dragon text files.
- Fixed a bug that caused duplicated descriptors in equations.
- Fixed a bug in fitting linear equations.
1.0.1.18 (Fri May 2 2008)
- Fixed a bug that caused access violation when pruning equations that contain nodes that have no coefficients.
- Several other minor bugfixes.
1.0.0.11 (Thu Apr 17 2008)
Version 1 is an almost complete rewrite of version 0.1.
- The genetic algorithm has been changed to a steady-state GA that resembles NSGA-II of Dep
and co-workers. It achieves elitism in an implicit manner.
- The parameter file syntax has changed.
- Each equation term can now have several adjustable (fittable) coefficients. This means that now it is possible to do Gaussian function approximation.
- Discarded much of the auxiliary testrun functions in the 0.1 version that added complication but provided nothing from the user's perspective.
- No parallel implementation.
0.1.1.4271 (Fri Mar 23 2007)
-
Fixed a bug in the Levenberg-Marquardt nonlinear least-squares routine.
-
Many fixes in the file reading routines.
0.1.0.4074
-
Fixed a bug in GOLPE data reading.
Revision 4040
-
Fixed a bug in SMILES generation.
Revision 3984 / 3987
-
Fixed a compiler inlining optimization related bug in Windows version.
-
Minor fixes in usability (more output when verbosity level is >=3).
Revision 3817
-
First public distribution.