baohan
/
scikit-learn
mirror of https://github.com/scikit-learn/scikit-learn.git
0
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
19,942 Commits 31 Branches 142 Tags 892 MiB
Commit Graph

47 Commits

Author SHA1 Message Date
Jan Hendrik Metzen 2649598c64 FIX Enforcing y to be numeric in GPR and fixing python3 issue with maps 2015-10-19 18:07:14 +02:00
Jan Hendrik Metzen 3957fc3b34 ENH GPR.log_marginal_likelihood() returns the current log-likelihood if no theta vector is provided 2015-10-19 18:07:14 +02:00
Jan Hendrik Metzen e2a340a4ee REF More meaningful names for hyperparameters of GP kernels 2015-10-19 18:07:14 +02:00
Jan Hendrik Metzen 2dd1f285b9 PEP8 Making pep8 and pyflakes happy 2015-10-19 18:07:14 +02:00
Jan Hendrik Metzen 213a6be540 FIX Fix message of ValueError for wrong shape of alpha in GPR 2015-10-19 18:07:14 +02:00
Jan Hendrik Metzen 1c4b23f47b MISC copy_X_train defaults to True in GPR and GPC 2015-10-19 18:07:14 +02:00
Jan Hendrik Metzen 65555b300b REF Renaming sigma_squared_n to alpha (consistency with Ridge) 2015-10-19 18:07:13 +02:00
Jan Hendrik Metzen c6db24e56f REF Several small changes based on @eickenberg's suggestions 2015-10-19 18:07:13 +02:00
Jan Hendrik Metzen c57aa64312 ENH X_train_ can be copied if requested 2015-10-19 18:07:13 +02:00
Jan Hendrik Metzen 92a6354aaf TST Fix GPC and GPR such that all tests are ok 2015-10-19 18:07:13 +02:00
Jan Hendrik Metzen 27b30166c3 FIX Several minor changes/fixes in gaussian_process module 2015-10-19 18:07:12 +02:00
Jan Hendrik Metzen a5e207dfb4 FIX GPR's random_state is cloned correctly 2015-10-19 18:07:12 +02:00
Jan Hendrik Metzen 0d34a34d5b FIX Ensuring that predicted variances are never negative 2015-10-19 18:07:12 +02:00
Jan Hendrik Metzen 74bbacbcc9 FIX Sampling theta_intial uniformly from log-bounds in GPR 2015-10-19 18:07:12 +02:00
Jan Hendrik Metzen 2839f66933 REFACTOR normalize_y in GPR does not modify standard-deviation 2015-10-19 18:07:12 +02:00
Jan Hendrik Metzen bcd4b72897 REFACTOR Remove theta_ attribute from GPR and GPC 
It can be accessed instead via kernel_.theta
 2015-10-19 18:07:11 +02:00
Jan Hendrik Metzen 4db5142bb8 REFACTOR GP kernel hyperparameters theta and their bounds are log-tranformed 
This representation of the search space  is more amenable for 
hyperparameter search, as hyperparameters like
length-scales naturally live on a log-scale.
 2015-10-19 18:07:11 +02:00
Jan Hendrik Metzen 897da0fd3f ENH GPR can use an externally defined optimizer for hyperparameter tuning 2015-10-19 18:07:11 +02:00
Jan Hendrik Metzen 76a1b16b75 FIX Not performing hyperparamter-tuning in GPR if there are no hyperparameters 2015-10-19 18:07:11 +02:00
Jan Hendrik Metzen 84b49f9c7a ENH GPR supports target value with more than one dimension 2015-10-19 18:07:11 +02:00
Jan Hendrik Metzen 66acd22f9e ENH GPR supports normalization of target values 2015-10-19 18:07:11 +02:00
Jan Hendrik Metzen cb9935b729 DOC Extending further the GP documentation 2015-10-19 18:07:11 +02:00
Jan Hendrik Metzen d26c355374 DOC Adding documentation for GP kernels 2015-10-19 18:07:11 +02:00
Jan Hendrik Metzen 7f8bd09322 ENH GaussianProcessRegressor supports multiple restarts of the optimizer 2015-10-19 18:07:10 +02:00
Jan Hendrik Metzen 02ab93e809 ENH GPR can use individual noise-levels for each datapoint 2015-10-19 18:07:10 +02:00
Jan Hendrik Metzen caa4196c3a DOC Revised docstrings of kernels.py 2015-10-19 18:07:10 +02:00
Jan Hendrik Metzen 7773707e0a DOC gpc module fully documented 2015-10-19 18:07:10 +02:00
Jan Hendrik Metzen 3073d7e395 DOC gpr module fully documented 2015-10-19 18:07:10 +02:00
Jan Hendrik Metzen 1c70e52dd6 ENH Adding diag(X) method to GP kernels. 
Using this method rather than np.diag(kernel(X)) or np.apply_along_axis
reduces memory-consumption and speeds up computation considerably
 2015-10-19 18:04:56 +02:00
Jan Hendrik Metzen 6af3ffc3c4 FIX Numpy downward-compatible way of computing sum over two axis 2015-10-19 18:04:56 +02:00
Jan Hendrik Metzen 40b1426499 FIX GaussianProcessRegressor inherits from RegressorMixin 2015-10-19 18:04:56 +02:00
Jan Hendrik Metzen 0a23d9e604 REFACTOR Compute standard deviation of predictive distribution at n points in O(n) 2015-10-19 18:04:56 +02:00
Jan Hendrik Metzen 141eb0b27c REFACTOR Rename GPR method sample to sample_y 2015-10-19 18:04:56 +02:00
Jan Hendrik Metzen 2feb6e02cf REFACTOR Replacing all calls to deepcopy() by clone() 2015-10-19 18:04:55 +02:00
Jan Hendrik Metzen 403b5f1412 REFACTOR REnamed kernel property params to theta 2015-10-19 18:04:55 +02:00
Jan Hendrik Metzen d0bc5dc35c FIX GPs do not modify kernel attribute but store optimized kernels in kernel_ 2015-10-19 18:04:55 +02:00
Jan Hendrik Metzen ef864824d0 ENH More efficient computation of lml-gradient in GPR using einsum 2015-10-19 18:04:55 +02:00
Jan Hendrik Metzen cb7c4243e8 MISC Reorganizing kernel module 2015-10-19 18:04:55 +02:00
Jan Hendrik Metzen d8405c2403 REFACTOR Hyperparameter optimization in GPs enabled explicitly by optimizer parameter 2015-10-19 18:04:55 +02:00
Jan Hendrik Metzen 722ddf197e FIX Dealing with singular kernel matrix (-inf log-marginal-likelihood) 2015-10-19 18:04:55 +02:00
Jan Hendrik Metzen e165fed1cc REFACTOR GaussianProcessClassification/Regression -> ...Classifier/Regressor 2015-10-19 18:04:55 +02:00
Jan Hendrik Metzen 03e35758f8 REFACTOR Merging kernel methods auto() and cross() in __call__() 2015-10-19 18:04:54 +02:00
Jan Hendrik Metzen 2ab097f0d3 TST Dealing with failing tests from test_common 2015-10-19 18:04:54 +02:00
Jan Hendrik Metzen a5dc3491ad MISC Revised gaussian_process package based on @eickenberg's suggestions 2015-10-19 18:04:54 +02:00
Jan Hendrik Metzen 1aa49b7a04 ENH GPR.predict allows returning standard-deviations of predictions 2015-10-19 18:04:54 +02:00
Jan Hendrik Metzen 515defe7a4 REFACTOR auto_correlation() -> auto(), cross_correlation() -> cross() 2015-10-19 18:04:54 +02:00
Jan Hendrik Metzen ca03d52bb7 ADD New implementation of GaussianProcessRegression 2015-10-19 18:04:54 +02:00