aboutsummaryrefslogtreecommitdiff
path: root/src
AgeCommit message (Collapse)AuthorFilesLines
2020-07-12Print training accuracyHEADmasterGertjan van den Burg2-5/+17
2019-06-24bugfix for kernel rbfGertjan van den Burg1-1/+1
2018-03-27Allow seeding for nonlinear GenSVMGertjan van den Burg2-25/+25
2018-03-27Add command line flag to set the maximum number of iterationsGertjan van den Burg1-0/+4
2018-03-27Bugfix for reading kernel parameters from training filesGertjan van den Burg1-3/+3
2018-03-27Add support for predicting after grid searchGertjan van den Burg2-18/+105
With this commit the gensvm_grid executable can now compute predictions with the best model found during the grid search. The test dataset is supplied through the training file, and a command line flag is added to support saving the predictions in an output file.
2018-03-27Major bugfix for nonlinear GenSVMGertjan van den Burg1-1/+4
Nonlinear GenSVM depends on the eigendecomposition of the kernel matrix. Mathematically, the Sigma vector in the code should contain the square root of the eigenvalues. Taking the square root was however neglected, which resulted in poor performance of nonlinear GenSVM. This is now fixed, which means that the performance of nonlinear GenSVM will be much better.
2017-12-13initialize model size parameters to zeroGertjan van den Burg1-0/+4
2017-02-23Allow setting of the random seed in the modelGertjan van den Burg5-7/+26
2017-02-21fix signed/unsigned comparisonGertjan van den Burg1-2/+1
2017-02-21Explicitly specify void argument to functions which takeGertjan van den Burg7-9/+6
no arguments.
2017-02-21Minor fixesGertjan van den Burg2-2/+12
2017-02-21Keep a status of trainingGertjan van den Burg2-2/+11
2017-02-21Keep track of elapsed iterations in trainingGertjan van den Burg2-0/+4
2017-02-17allow max_iter to be set in the modelGertjan van den Burg4-9/+11
2017-02-17Remove kernelparam array in favour of explicit kernel parametersGertjan van den Burg7-172/+83
This simplifies a lot of the code and will make it easier to link to other languages.
2017-02-17minor code clarificationGertjan van den Burg1-1/+1
2017-01-02documentation fixesGertjan van den Burg1-3/+3
2017-01-02fix for string versionGertjan van den Burg3-3/+3
2016-12-08add specification of libsvm data formatGertjan van den Burg1-1/+2
2016-12-08Add copyright and license notice to execsGertjan van den Burg2-2/+8
2016-12-08Add range check for some variablesGertjan van den Burg1-2/+22
2016-12-07allow datasets to be stored in libsvm/svmlight formatGertjan van den Burg5-15/+428
2016-12-07moved check for class labels to seperate moduleGertjan van den Burg4-41/+99
2016-12-07throw warning when using sparse matrices with kernelsGertjan van den Burg2-0/+31
2016-12-07make test for sparsity the same everywhereGertjan van den Burg1-4/+3
2016-12-07only print kernel computation method with nonlinear kernelsGertjan van den Burg1-2/+4
2016-12-07print warning when a test dataset is supplied in training fileGertjan van den Burg1-2/+6
2016-12-07add ifndef and GENSVM_ before definesGertjan van den Burg7-42/+48
2016-12-07switch some integer variables to long for cross platform size guaranteesGertjan van den Burg7-17/+17
2016-12-06document undocumented elementsGertjan van den Burg7-59/+187
2016-12-06add unit tests for gensvm_trainGertjan van den Burg1-1/+2
2016-12-05fix missing test in gensvm_optimize and expose doublesortGertjan van den Burg2-5/+4
2016-12-05further unit tests for kernel moduleGertjan van den Burg2-36/+45
2016-12-05Tests and documentation kernel moduleGertjan van den Burg1-81/+167
2016-12-05Improvements and tests for tasks and gridsearchGertjan van den Burg2-5/+65
2016-11-07compute ZAZ in blocks to increase numerical precisionGertjan van den Burg2-23/+62
2016-11-03prepare for gridsearch unit testingGertjan van den Burg6-380/+522
2016-10-24update copyright informationGertjan van den Burg26-85/+519
2016-10-18fix problems with dense ZV computation when data->n < model->nGertjan van den Burg1-1/+2
2016-10-17minor documentation fixesGertjan van den Burg2-6/+5
2016-10-17refactor gensvm_pred to gensvm_predictGertjan van den Burg2-2/+2
2016-10-17Update predictions to work with sparse matricesGertjan van den Burg3-59/+118
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
2016-10-17Add functionality for cv_util for sparse matricesGertjan van den Burg1-4/+168
2016-10-17Create debug function for printing GenSparse structsGertjan van den Burg1-1/+44
2016-10-14remove temporary test seed settingGertjan van den Burg1-4/+1
2016-10-14documentation fixesGertjan van den Burg7-14/+41
2016-10-14add sparse matrices to GenSVM and reorganize update functionalityGertjan van den Burg7-499/+877
2016-10-06documentation fixesGertjan van den Burg2-2/+2
2016-10-06make use of the fact that the first element of z_i is always 1Gertjan van den Burg1-3/+8
generated by cgit v1.2.3 (git 2.25.1) at 2026-01-09 08:56:51 +0000