- Added a more generous epsilon procedure for convenience.
- Added support for user-provided kernel functions.
- Added a utility for building a sparse matrix from a function on the data.
- (Re)added separate TMDmap class wrapping base diffusion map class to allow for easier construction of TMDmaps.
- Added ability to explicitly provide the sampled density for q^alpha normalization.
- Added Variable Bandwidth Diffusion Maps.
- Added a new out-of-sample extension method that should work for variable bandwidth methods.
- Changed from exp^(-d^2) to exp^(-d^2/4) convention.
- Moved weight functionality into a function provided on initialization, rather than input values, and added a helper function that allows values to be read from a lookup table.
- Improved the Diffusion Map test suite.
- Moved out-of-sample routines into separate functions.
- Moved matrix symmetrization into newly made utility file.
- Adjusted constructor for the diffusion map to take the kernel object directly.
- Fixed bug where weight matrices were not included for out of sample extension.
- Moved to MIT license.
- Fixed setup.py issues.
- Added base functionality to the code.