- Fixed an error in
do.lapeig() thanks to Frithjof
Kruggel (UC Irvine).
- Removed dependence on
Rcsdp package.
- Fixed potential errors with some examples.
- Removed arbitrary regularization in
do.dne().
- Change of authorship.
- Change of some estimation methods to return implicit variables.
do.dppca() added : dual probabilistic principal
component analysis.
- Change of maintainer’s contact.
- Some functions are converted to pure C++.
- Fixed an error in
do.olpp() thanks to Nicholas
Spyrison.
- Replaced nearest neighbor library.
- Methods added;
do.mmds(), do.phate().
- Bibliographic errors fixed.
- Methods for feature selection
do.wdfs(),
do.uwdfs(), do.procrustes(), and
do.mifs() are added.
README now shows numbers of currently available
functions for DR and IDE.
- Fixed memory leaks in
do.sne() and
do.tsne().
do.lsls() added as a supervised feature selection
method.
- README contains minimal examples for both dimension reduction and
estimation.
- Porting to pure C++ implementations started, gaining computational
efficiency.
do.lmds function is fixed for its discrepancy in nested
structure.
- NEWS reformatted and package website is now
available.
- Dependency on R package ADMM is removed.
Major changes
- LDA solves trace ratio problem directly.
- Many of dependencies are removed.
- 133 dimension reduction methods available.
- 17 intrinsic dimension estimation methods available.
Bug fixes
- Error fixed in
do.lscore function (thanks to Jordan
Lin).