| Age | Commit message (Collapse) | Author | Files | Lines | |
|---|---|---|---|---|---|
| 2020-07-12 | Print training accuracyHEADmaster | Gertjan van den Burg | 1 | -0/+1 | |
| 2016-10-24 | update copyright information | Gertjan van den Burg | 1 | -2/+20 | |
| 2016-10-17 | Update predictions to work with sparse matrices | Gertjan van den Burg | 1 | -0/+1 | |
| This is done by pulling the Z*V routines from the gensvm_optimize file to a seperate file, since they are shared by prediction and get_loss | |||||
| 2016-10-14 | add sparse matrices to GenSVM and reorganize update functionality | Gertjan van den Burg | 1 | -19/+6 | |
| 2016-10-06 | documentation fixes | Gertjan van den Burg | 1 | -1/+1 | |
| 2016-10-06 | Switch to using dsyrk instead of dsyr for speed. | Gertjan van den Burg | 1 | -3/+4 | |
| Also added a workspace (GenWork) structure for to hold working matrices for the gensvm_get_update() and gensvm_get_loss() functions | |||||
| 2016-09-30 | rewrite of the update function to use only rank 1 operations where possible | Gertjan van den Burg | 1 | -5/+3 | |
| 2016-09-30 | Remove category matrix from implementation | Gertjan van den Burg | 1 | -4/+4 | |
| It's a remnant from the Matlab implementation, where it actually makes sense to use. Here it becomes a significant memory burden for large datasets, and doesn't provide any speedup | |||||
| 2016-09-30 | Break up update function and add unit tests for parts | Gertjan van den Burg | 1 | -1/+12 | |
| 2016-05-16 | create a single training function for easy external access | Gertjan van den Burg | 1 | -2/+2 | |
| 2016-05-16 | major refactor of the code | Gertjan van den Burg | 1 | -0/+36 | |
