Version 3.0

From the change log:

• Removed class for polynomial regression (PolynomialRegression).
• Removed class for radial basis function approximation (RBFApproximant)
• Updated MATLAB and Python examples
• Added a new builder class for B-splines (BSpline::Builder)
• Changed default architecture to x86-64
• Added support for equidistant knot vectors (still experimental)
• Added support for selecting the number of basis functions when constructing a B-spline
• Added support for building splines with a regularization term (as an alternative to the P-spline)
• Replaced all assertions with exceptions
• Reworked MATLAB, Python, and C interfaces
• Changed to Eigen version 3.2.8
• Updated documentation

Please note that as this is a new major release (2 -> 3), the interface has been changed and all existing code will with all likelihood be broken by this release.

OS X users: We currently offer no precompiled binaries for OS X, so you will have to compile SPLINTER yourself. Sorry for the inconvenience!

Version 2.0

List of changes (from the change log):

• Automatic knot vector selection using moving average
• Added PolynomialRegression (including MatLab interface)
• Added saving and loading of PSpline, RadialBasisFunction and PolynomialRegression
• Reworked serialization (~5% faster serialization now)
• Added loading of serialized objects in MatLab interface: b = BSpline('saved_bspline.bspline')
• Added loading and saving of DataTable to MatLab interface
• Refactored MatLab interface backend (no end user visible changes)
• Integrated the Catch testing framework and added extensive testing of approximation, serialization and datatable set operation features
• Added set operations (union and complement) to DataTable (as operator+ and operator-)
• Added script (scripts/build_release.sh) for easier compilation (especially on Windows with MSVC)
• Improved compilation documentation
• Renamed API macro to SPLINTER_API to avoid name collisions with users namespace
• Fixed BSpline hessian not being symmetric
• Added evalJacobian implementations for PolynomialRegression and RadialBasisFunctions
• Added standard C++11 types to Approximant interface
• Refactored B-spline code.
• Added Python interface to the library
• Added batch evaluation to MatLab interface
• Renamed Approximant subclasses class names, example: BSpline -> BSplineApproximant

Version 1.3

Please consult the change log to see what changes this release brings.

Note the interface change in the BSpline class: It now accepts Linear, Quadratic, Cubic and Quartic types.

Version 1.2 new library name

This release is basically the same as the previous, except for the new name.

Version 1.2

Added saving and loading of bsplines and datatables in binary form.

Note: The Windows binaries might have been compiled with less optimizations than desired (-O2 vs -O3). There will hopefully be an update in the future with binaries that are compiled with all optimizations.

Initial release

Pre-compiled binaries for the initial release of the library.