Releases: oysstu/LabVIEW-libsvm
Releases · oysstu/LabVIEW-libsvm
v1.1
v1.0
v0.7-beta
- Added Linux support. Tested on mint 17.2 (x64) and CentOS 7 (x64) with both LabVIEW x86 / x86_64.
- OpenG dependencies limited to the exact libraries needed.
- Additional runtime input validation checks in library added.
- Added model saving/loading for liblinear.
v0.6-beta
- Fixed critical bugs in the liblinear wrapper for some solver types
- Added two example VIs for the liblinear wrapper
- Fixed bugs in the grid-search visualization
- Removed scaling of labels (don't see the need to have this)
- Added VI that generates a stratified validation/training set split (randomized)
- Restricted backwards compatibility to LabVIEW 2013 (aiming at supporting the last three versions)
v0.5-beta
- Added dense implementation of SVM, which can boost performance depending on the dataset
- Added VI for parallelized gridsearch of RBF parameters (C, gamma)
- DLLs are now compiled statically to reduce the number of dependencies, also LabVIEW will now automatically include them when building an application.
- Examples for the dense implementation drops the logging functionality, which makes them easier to follow.
- Various minor bugfixes both in the DLL and LabVIEW.
v0.4-alpha
- Reentrancy set to shared clone on cross-validation, training and prediction to facilitate parallel processing.
- Tool for computing confusion matrix for binary classification results
- Tool for importing a spreadsheet file to svm_problem format (similar to http://www.csie.ntu.edu.tw/~cjlin/libsvm/faq.html#f307)
- Added a simpler example for training a C_SVC classifier (without library logging).
- Minor interface change on linear scaling VIs
Bugfix release
Fixed crashes that occured while using 32bit LabVIEW
Added file functionalty and linear data scaling
v0.2-alpha Fixed bug: -inf +inf was swapped