Commit Graph
21273 Commits
Author SHA1 Message Date
favre49 a14a5259a0 Merge pull request #2006 from rcurtin/dt-test-fix
Fix `DecisionTreeTest/PerfectTrainingSet`
2019-09-11 19:15:26 +05:30
Ryan Curtin cf71ecef94 Merge pull request #2011 from rcurtin/ensmallen-2-fixes
ensmallen 2 compatibility fixes
2019-09-10 17:29:14 -04:00
Ryan Curtin 65878a16f4 Merge pull request #2009 from rcurtin/dt-main-test-fixes
Fix occasional `DecisionTreeMainTest` bugs
2019-09-10 17:26:58 -04:00
Ryan Curtin 388014a3e0 Merge pull request #2004 from rcurtin/gmm-train-test-fix
Fix intermittent failures in GMMTrainMainTest
2019-09-10 10:38:04 -04:00
Ryan Curtin 8f3d166bb6 Increase travis build timeout. 2019-09-08 15:21:53 -04:00
Ryan Curtin d716e312f6 Merge pull request #2010 from rcurtin/ssrbm-tolerance-fix
Widen tolerance for ssRBMClassificationTest
2019-09-08 13:55:29 -04:00
Ryan Curtin f4846ad8bd Also initialize action member in copy/move constructors. 2019-09-08 13:44:02 -04:00
Ryan Curtin fecb360131 Merge pull request #2007 from KimSangYeon-DGU/knn_test
Fix `KNNMainTest/KNNAllTreeTypesTest`
2019-09-08 11:48:56 -04:00
Ryan Curtin c475d9aa10 Handle both ensmallen 1 and 2 API. 2019-09-08 02:19:39 -04:00
Ryan Curtin f5124058a2 Widen tolerance just a bit. 2019-09-07 22:24:09 -04:00
Ryan Curtin 90fe0b6a80 Cleanups for DecisionTreeMainTest. 2019-09-07 14:44:57 -04:00
Ryan Curtin 8a168d1b26 Remove unnecessary column from braziltourism ARFF files. 2019-09-07 14:34:47 -04:00
KimSangYeon-DGU be67ebd912 Increase k from 10 to 15 to make this test robust. 2019-09-07 14:06:22 +09:00
Ryan Curtin 7ade459e74 Fix strange bug. How did that get there? 2019-09-06 15:59:09 -04:00
Ryan Curtin 76583b96d6 Refactor and clean up GMM training main tests. 2019-09-05 22:30:00 -04:00
Ryan Curtin b739b0ff04 Add --kmeans_max_iterations. 2019-09-05 22:29:53 -04:00
Ryan Curtin 33e1599cfe Add utility to set a specific random seed. 2019-09-05 22:29:38 -04:00
Ryan Curtin 444559d71e Merge pull request #1995 from rcurtin/rf-main-test-fix
Fix `RandomForestMainTest` random failures
2019-09-04 17:08:41 -04:00
Ryan Curtin bb979c43c5 Merge pull request #2001 from rcurtin/lsh-test-fix
Fix very occasional LSH test failures
2019-09-04 17:08:21 -04:00
Ryan Curtin 8230031ae4 Merge pull request #2002 from rcurtin/lsvm-test-fix
Increase Linear SVM test tolerance slightly
2019-09-04 17:07:53 -04:00
Shikhar Jaiswal e58bea3d7d Merge pull request #1913 from saksham189/mini
Add MiniBatchDiscrimination Layer
2019-09-04 05:33:38 +05:30
Ryan Curtin 53cdd0b291 Merge pull request #1996 from favre49/RevertRL
Revert Action.action to be a double array
2019-09-03 14:37:20 -04:00
Ryan Curtin 87694a394c Merge pull request #1998 from zoq/SimplePendulumTestFix
Fix Pendulum task max steps test
2019-09-03 14:36:13 -04:00
Ryan Curtin f55bee736d Merge pull request #1999 from zoq/RewardClippedAcrobotWithDQNFix
DQN reward-clipping test adjustments.
2019-09-03 14:35:52 -04:00
Ryan Curtin 30ebf5d9c0 Merge pull request #2000 from zoq/SerializationNormalizationTest
double-precision serialization is not lossless
2019-09-03 14:35:13 -04:00
Ryan Curtin 976e1afe8f Increase tolerance slightly.
Sometimes we can get unlucky and L-BFGS won't optimize quite right.
2019-09-03 09:42:20 -04:00
Ryan Curtin 2fe65883f9 Add noise to test to prevent very occasional failures. 2019-09-02 18:56:54 -04:00
Marcus Edel d867c7099f Address code review comments (documentation). 2019-09-01 21:48:29 +02:00
Marcus Edel 4e8bb575b6 On macOS double-precision serialization is not lossless, so switch to float. 2019-09-01 19:48:06 +02:00
Rahul Prabhu 7160d5cfa2 Fixed HISTORY 2019-09-01 19:20:11 +05:30
Marcus Edel e6cd8b6217 Run the DQN reward-clipping test multiple times to account for minimal model size and short test time. 2019-09-01 15:32:50 +02:00
Marcus Edel f50fc306c3 There is no guarantee that we will reach the maximum number of steps with a random action, so we have account for that. 2019-09-01 13:53:21 +02:00
Rahul Prabhu e5a4c40afb Reverted action.action to a double array 2019-09-01 10:52:53 +05:30
Ryan Curtin 9e5e9f9451 Add --maximum_depth test. 2019-08-31 22:54:52 -04:00
Ryan Curtin 2f84895509 Compare the leaf sizes and the number of trees directly.
This is better than the accuracy test, which can sometimes fail.
2019-08-31 22:49:52 -04:00
Marcus Edel d4c53ec546 Merge pull request #1992 from conradsnicta/spmat-zero-write-fix
More fixes to avoid writing zero via sparse matrix iterator.
2019-08-31 19:25:11 +02:00
conrad 79f175aece avoid writing zero via sparse matrix iterator 2019-08-31 15:49:51 +10:00
Ryan Curtin 6aee9ecbf3 Merge branch 'conradsnicta-master' 2019-08-29 18:15:25 -04:00
Ryan Curtin 70c531dc2d Update history. 2019-08-29 18:15:16 -04:00
Ryan Curtin 0c7e9decc8 Fix style issues. 2019-08-29 18:14:49 -04:00
Ryan Curtin c2a0567d08 Merge branch 'master' of https://github.com/conradsnicta/mlpack into conradsnicta-master 2019-08-29 18:14:31 -04:00
Ryan Curtin 131b376d06 Merge pull request #1984 from robertohueso/reclaim-kde-error
[GSoC] Reclaim unused KDE error tolerance
2019-08-29 18:13:07 -04:00
conrad 23204880a3 added TODO to consider using .transform() instead of iterators 2019-08-29 00:44:49 +10:00
conrad 1a89e79574 avoid writing zero via sparse matrix iterator 2019-08-29 00:41:49 +10:00
Roberto Hueso Gomez d35b2f7bea Improve KDETest style 2019-08-26 19:47:42 +02:00
Roberto Hueso Gomez 0ca13852e9 Add StandardCoverTreeEpanechnikovKDETest 2019-08-26 19:42:17 +02:00
Roberto Hueso Gomez bb4955fbb1 Add GaussianCoverSingleKDETest 2019-08-26 19:34:46 +02:00
Roberto Hueso Gomez d232ff3455 Reduce KDE tests sizes to save time 2019-08-26 19:27:22 +02:00
Roberto Hueso Gomez ab31584138 Improve KDE docs 2019-08-26 06:18:52 +02:00
Roberto Hueso Gomez 2ff04907a4 Remove centroid from KDEStat 2019-08-26 06:06:10 +02:00