[scikit-learn] annotated tag 0.1-beta created (now 602f76f)
Andreas Tille
tille at debian.org
Wed Dec 28 13:10:47 UTC 2016
This is an automated email from the git hooks/post-receive script.
tille pushed a change to annotated tag 0.1-beta
in repository scikit-learn.
at 602f76f (tag)
tagging e6989efd71a2adddd03979d1fe7a2e82e37ea51f (commit)
tagged by Yaroslav Halchenko
on Thu Mar 25 08:46:37 2010 -0400
- Log -----------------------------------------------------------------
Tagging 0.1-beta release
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
iEYEABECAAYFAkurWy0ACgkQjRFFY3XAJMgfowCgzx8fTspWIfzuu/0Xk23iiKgv
sNAAn0Xv0IHP8hF2+kFVdxVaHF8m62Bb
=nTy/
-----END PGP SIGNATURE-----
Fabian Pedregosa (382):
Initial repository layout
Initial commit for learn, a set of algorithms and tools for machine learing in python
Adding oldfaithful to datasets
Adding Iris data from UCI ML db
Adding pendigits from UCI ML DB
Add __init__ file to iris dataset module
Add heart data
Add German credit dataset
Add __init__ file to german and heart datasets
Split german data between features and label
Add learn.preprocessing, for data pre processing
Add a basic readme to learn.datasets package
Flatten data in Iris dataset
Add a misc subpackage in datasets for tools useful to several datasets
Flatten oldfaithful
flatten german and pendigits
Add a scaler object + tests.
Add function to handle Nan data for scaling.
First code commit.
From: fred.mailhot <fred.mailhot at cb17146a-f446-4be1-a4f7-bd7c5bb65646>
Adding data dir with data files.
- Got least-squares optimization working
Implementing error/gradient functions more carefully. Borrowing heavily from Netlab.
Extensive refactoring of mlp (still convergence probs with some optimization algs). Started work on srn, should commit a good deal more tonight (want to have a working srn with at least one training alg before the weekend).
Minor changes to srn.py. Plans for future dev included in docstrings & comments.
New start with RBF net. Will finish this Sunday.
Heavily refactored MLP. Only leastsq used for optimization.
Refactoring SRN based on MLP. Simple 1-step backprop only.
data files for testing
more data files for testing
Heavy refactoring. These are more or less complete now. Small additions remain to make the interface more obvious for users.
Updates/refactorings for mlp and srn, addition of rbf.
Removed two data files.
Updated __init__
Easier way to select random sample of data for RBF centers.
HTML files for {mlp/rbf/srn}.py generated by pydoc uploaded.
Add axis arguments to functions that changed in NumPy 1.0b2
initial checkin of ga library.
much clean up of module paths after files were moved.
initial attempt at including Numeric as a sub-package
attempt at fixing the chaos created by my last checkin.
fixed several package related issues with from x import *
added from fastumath import * to all sub-modules in scipy (besides Numeric)
Fixed a time.time function replacement error that Prabhu found.
adding setup_xxx.py files for the new setup.py infrastructure.
Fixed import of stats module in ga.
Moved shelve to io. Cleaned up __init__ files.
Changed references to fastumath to scipy_base.fastumath
Changed == None to is None
Converted module to use 'import' directly, instead of exec 'import...'
Carried out major unification of xxx/setup_xxx.py files. Discussion: some modules contain setup.py files that repeat the functionality of the corresponding setup_xxx.py files. Are there any objections if setup.py and setup_xxx.py will be merged into setup_xxx.py and setup.py files will be removed from CVS?
1) Finished applying ppimport hooks to scipy.
* Made changes to bring ga up to speed with latest stats module.
Replaced execfile statements with simplified import statements in order to avoid problems with Windows installer tools.
New import hooks are applied to cow,ga,sparse. Clean up.
Fixed import problem with importing __pre___init__
Converted usage of whrandom to random.
More changes to get rid of whrandom usage.
Fixed issue 216
Converted scipy to use numpy. Several tests fail but are easily fixed.
Fixed old imports (ticket #36)
Run reindent.py (script distributed with Python) over the source to remove extraneous whitespace
remove unused imports
remove duplicate definition of __float__
Add axis arguments to functions that changed in NumPy 1.0b2
Initial commit David Cournapeau <david at ar.media.kyoto-u.ac.jp> | 2006-06-08 17:59:58 +0900 (Thu, 08 Jun 2006)
Put files into a package David Cournapeau <david at ar.media.kyoto-u.ac.jp> | 2006-07-12 13:52:02 +0900 (Wed, 12 Jul 2006)
Package uses distutil David Cournapeau <david at ar.media.kyoto-u.ac.jp> | 2006-07-13 15:20:08 +0900 (Thu, 13 Jul 2006)
Merge with main branch David Cournapeau <david at ar.media.kyoto-u.ac.jp> | 2006-07-13 19:35:51 +0900 (Thu, 13 Jul 2006)
Add Changelog, put version to 0.3 David Cournapeau <david at ar.media.kyoto-u.ac.jp> | 2006-07-13 19:51:04 +0900 (Thu, 13 Jul 2006)
Refactoring of EM into classes David Cournapeau <david at ar.media.kyoto-u.ac.jp> | 2006-07-14 16:26:45 +0900 (Fri, 14 Jul 2006)
Add scripts for benchmarking, minor corrections David Cournapeau <david at ar.media.kyoto-u.ac.jp> | 2006-07-14 19:28:13 +0900 (Fri, 14 Jul 2006)
Push to version 0.4.2 David Cournapeau <david at ar.media.kyoto-u.ac.jp> | 2006-08-04 19:48:19 +0900 (Fri, 04 Aug 2006)
Version 0.5, add custom confidence ellipsoids David Cournapeau <david at ar.media.kyoto-u.ac.jp> | 2006-08-07 18:46:57 +0900 (Mon, 07 Aug 2006)
Adapt to numpy 1.0b3SVN David Cournapeau <david at ar.media.kyoto-u.ac.jp> | 2006-08-17 18:48:51 +0900 (Thu, 17 Aug 2006)
correct bug with GMM.init_method David Cournapeau <david at ar.media.kyoto-u.ac.jp> | 2006-08-24 19:44:42 +0900 (Thu, 24 Aug 2006)
revert to custom kmean David Cournapeau <david at ar.media.kyoto-u.ac.jp> | 2006-08-24 20:12:27 +0900 (Thu, 24 Aug 2006)
Various cleanup, include ctypes version of diag gden; update profile function to include ctypes version David Cournapeau <david at ar.media.kyoto-u.ac.jp> | 2006-08-28 22:03:28 +0900 (Mon, 28 Aug 2006)
- Add a plot1d method to GM class
- Improve plot1d method to be more useful
MSG David Cournapeau <david at ar.media.kyoto-u.ac.jp> | 2006-10-03 18:29:50 +0900 (Tue, 03 Oct 2006)
Bump to 0.5.3 David Cournapeau <david at ar.media.kyoto-u.ac.jp> | 2006-10-03 18:34:42 +0900 (Tue, 03 Oct 2006)
Add preliminary online functions (does not work yet) David Cournapeau <david at ar.media.kyoto-u.ac.jp> | 2006-10-06 20:03:59 +0900 (Fri, 06 Oct 2006)
Change of layout for inclusion in scipy David Cournapeau <david at ar.media.kyoto-u.ac.jp> | 2006-10-12 21:19:17 +0900 (Thu, 12 Oct 2006)
Change of layout for scipy (2) David Cournapeau <david at ar.media.kyoto-u.ac.jp> | 2006-10-12 21:21:33 +0900 (Thu, 12 Oct 2006)
Last pyem layout cleanup to finalize inclusion into scipy trunk
Add svm and pyem package in the comments of setup.py; correct bugs in the online EM script
* Convert the online_em.py script to a class, OnGMM * Add tests for OnGMM * Small bugfixes in _c_densitites (convert general exception to ImportException when wrong ctypes version found), setup (remove the name attribute to avoid warning when installing)
* pyem: added GM.bic function to compute Bayesian Information Criterion for automatic model selection + various docstrings fixes
* pyem: trivial change of API for OGMM
* sandbox.pyem: fix a bug which prevents full covariance from being used + tests to avoid this issue from being unnoticed
* bump to 0.5.6 * various cosmetic changes
Various bug fixes (see Changelog)
Add densities2.py for preliminary axis support and add specialized class for online EM in 1d for a 10-100x speed incread
Add a script to profile online em
Move documentation files in doc repository. * all examples lie in doc/examples. * A Makefile to generate the pdf from the rest files and python examples is now included.
Put examples directory at the top level, so that they be easily imported for tests.
Remove kmean as scipy.cluster.vq.kmeans2 does everything we need now
More benchmarking for basic operations in row vs col
Refactor 1d computation for plotting
Add some profiling scripts to compare likelihood computation with matlab.
Add faithful data in data.
Add function to plot density contours in GM.
Add proper license in data, correct typo (double copyright)
Polish contour functions, so that choosing the dimension of projection works.
Fail nicely when call wrong plot function (plot1d for multinomial, plot for 1d models).
Add example of pdf estimation with EM
Add data as a proper submodule
Clean up densities.py code, set docstrings to rest
Set def arguments to immutable to avoid nasty side effect.
Heavy liftup of the code + docstrings.
Add special initialization method for mixture models for testing purpose.
* Correct bogus GM._get_va which caused bogus isodensity plot + test * Support for plain matrix in GM.check_state
Add a plotting example.
Remote outdated test script.
Add basic tests for EM, 1d, 2d, full and diag mode
Reenable tests I forgot to uncomment in gmm_em tests
Convert check calls to test calls in tests, for future convertion to setuptools
Add tests for pdf computation in log domain (1st step for logsumexp trick support)
Add logsumexp function + tests. Not used in the code yet, though
Add function to compute log responsabilities with logsumexp.
Fix importing datasets in pyem/test
More fix for broken tests in gmm_em
Trivial fix for typo in pyem tests.
Refactor update step for EM (split diag and full case in subfunction)
add pendigits data from UCI machine learning website. Will be useful for testing regularization.
Add pendigits as a subpackage of data for distutils.
Change named arg varmode to mode for consistency, and add an argument for number of contours in density on grid
Update docstring of logsumexp.
Add (crude) regularized EM
Add support for EM in log domain + tests
Add utils function to easily load pendigits data
Add iris data from UCI ML database
Remove datasets (include them in sckits.learn instead)
Add curry class to misc tools
Remove data dir from subpackages in setup
Clean up some examples, and add an example for regularized EM
Forgot to add the file implementing regularized example...
Use log pdf when possible in plot functions
Clean up code for 1d plotting.
Clean up for densities.py
More clean up
Add discriminant analysis example
Update doc
Add forgotten picture necessary for generating the doc
Significantly improve speed of gauss_den
More speed improvements for full matrices case
pylibsvm SoC project getting under way.
libsvm 2.82 sources.
Basic classification, regression and one-class with new API.
Fixed tests.
Started on documentation.
Separated out kernels. Added facility for custom kernels.
Fixed RBF kernel.
New design to hide libsvm data structure from the users.
minor
Build a proper libsvm DLL on Windows.
Code integrated into NumPy itself.
Datasets.
Basic tests for datasets.
Minor changes due to code moving to Numpy.
Fix kernels.
Training of regression model works.
Predict stuff. Added some docs.
Classification test and some minor fixes.
Regression test. Workaround for possible ctypes bug. Probability output for classification models.
Workaround for bug in ctypes 0.9.9.6.
Extend one-class estimation a bit.
Cross-validation for classification and regression.
SConstruct file to build shared library until numpy.distutils gets sorted out.
Tweak so that test passes with GCC.
First attempt at a setup.py.
Howto for getting the test suite to run.
Minor updates.
Let ctypes handle references to the contents of the svm_problem.
'randn' is no longer in top level namespace. Call N.random.randn.
Simplify libsvm wrapper and make tests run without having to set PYTHONPATH.
Dataset for making precomputed kernels.
Precompute with any kernel.
Working on support for precomputed kernels.
Reformat code to conform to PEP 8.
Dataset for dealing with precomputed kernels.
Precomputed model training.
More cleanups to conform to PEP 8.
Factored out prediction logic so that we can support precomputed and custom kernels.
minor
Prediction refactoring in progress.
Predict without calling libsvm.
Code cleanup. Disabled tests that are failing due to NumPy defect.
More tests.
Collapse predictor classes to yield a nicer API.
Fix libsvm prediction with precomputed kernels.
Make sure custom kernels are only used with precomputed datasets.
Test for classification with precomputed kernel.
Don't train probability models by default.
Use .ctypes.data_as instead of ctypes.cast.
Use release build linker options.
Extend classification tests.
Enable tests that exposed NumPy defect.
Improve one-class tests.
Extend tests to check nu-SVR.
Classification in Python in preparation for some optimization.
Model compaction.
Move zipping inside the datasets.
Benchmark with large test dataset.
Refactoring kernels to work better in sparse and dense cases.
Fix a bug in ellipses of confidence computation for mixture + adapt to new datasets layout in examples
test
undo fix
resyncing with scipy version
Move former scipy.sandbox modules into a separate package machine
Modify setup to that svm and pyem are built from their setup
Enable scikits.learn.utils test through setuptools --test-module mechanism.
Adapt tests + module import in pyem such as they work with setuptools test command
Add test and test_suite function to svm module
Ignore junk files in the whole learn tree, make machine a proper subpackage by adding an __init__.py file
Make svm/examples a package so that its files are copied when installing
Prototype of attributes and class selection, based on introspection
Do not detect subpackages using find_packages from setuptools, but uses hiearchical setup.py ala scipy instead. Should solve #24
Set correct name for datasets packages (data->datasets) in the setup, and trivial changes in datasets info message
Changing pyem to em
Change examples and tests following the change of package name pyem->em
Fixed online em tests
Split pendigits into training and testing datasets, and convert return value of load to the package conventions.
Remove pyc junk.
Adapt examples using pendigits to the new format.
Put back changes which somewhat were not commited before for setup in em sub package
First version of arffreader. Seems to read correctly most headers. Data not parsed yet
Non working arff reader
Working arffreader
Printing a small summary for arffreader
More thoughts on arffreader
Handling missing values (?)
Simplify Kernal Api names
Finish api names simplification: no more Kernel suffix, no more LibSvm prefix
Add partial evaluation tool to replace functools module for python < 2.5
Make common a module
Add common as a module to install in setup
arffreader can now parse all numeric data of UCI :)
Catch error while parsing arff file, and raise ParseError for expected problems
Cleaning a bit the arffreader module
Add a small java program to parse arff using weka (meant to be used for testing our arff implementation)
Abstract away all meta data returned in user functions into a MetaData class
Add scripts for quick and dirty test from java
Adding dataset proposal
Added misc to setup.py
added a data set with missing entries.
Added the narr module to load NCEP NARR data sets (reanalysis of around 190 variables). Depends on pydap.
First manifold_learning_toolkit commit : not everything may work
Small update for the dimensionality reduction folder
Small argument mistake
Some bug fixes, some name changes, ...
Remove fake pyem package.
Remove the arff reader: an improved version is available in scipy.io, under the name loadarff.
Added some missing files
Added a simple compression example
Fixed some additional arguments and modules issues
Remove pyem from setup.py
Changed neighbo(o)rs name
Fixed some extensions bugs
Disable manifold learning, because it does not build.
Updates
Setup.py should be more robust now
Updated the classes according to the online tutorial
Small correction
refs #60: for gcc < 4.1, SFINAE seems to be bogus, so iteration copies are deactivated by a macro.
Rename NumpyTestCase to TestCase.
Remove path madness, which is deprecated anyway.
Do not use set_package_path anymore in machine.em tests.
Fix typos in tests.
Remove deprecated test runned for em machine.
Add online_gmm_em method.
Add compute factor.
Start working on em2, an backward incompatible, more scalable rewrite of em.
Start working on new GM class.
Be consistent between w and weights.
Add changelog.
Add a README for em2.
Trailing spaces.
Add logsumexp wrapper around cython implementation.
Implement logsumexp in cython.
implement pdf method for GM
Implement pdf for GM.
Do not execute example statements at import time.
Implement sample method for GM.
Postpone learn import.
Add a GMM class.
Add test code.
Start experimenting with accumulate API for EM.
Add kmeans-based init for w, mu, va.
Start EM imp.
1e-15 too stringent.
Remove GMM class.
Use new classes for new EM.
Whitespace.
Add em2 to learn scikits.
Implement one component likelihood function.
Add (buggy) cython implementation for one comp likelihood function.
Add tests for logsumexp and mnormalik.
Add basic test for EM.
Update README.
Remote hint parameter from EM ctor.
Fix utils subpackage __init__; used old numpy testing framework.
Raise exception in normalik, since it is buggy.
Fix import.
Add test for logresp.
Revert "Raise exception in normalik, since it is buggy."
Make a frame per frame inline logsumexp to reuse it.
Use a frame/frame inline quadform to reuse it.
More tests for logresp (python version).
Start working on cython version of logresp.
Split tests for logresp python vs cython.
cython logresp now works.
Rename methods of the Scaler class.
Do not install em2 by default, not mature enough yet.
Moved from old openopt to the new optimization scikit
Deleted Boost dependency (should OK now)
BUG: port em tests to new numpy >= 1.2 infrastructure.
Added README file.
Use nosetest for testing.
Disable test for this dataset
Add .gitignore
affread is now part of scipy.io.arff
Add __init__.py file to the test module.
Fix import paths in manifold_learning.*
Remove old testing framework
Web page sphinx infrastructure.
Update README file.
Fix: Add std namespace in order to use std::isnan.
Build extensions inplace.
Update README
Added COPYING file with the license terms.
Update README.
Use unittest.TestCase instead of numpy.testing.TestCase.
Added funding orgs to web/funding.rst
Add warning to web page.
Add example to Neighbors.kneighbors in machine.manifold_learning.neighbors
Revert commit 229 (inclusion of namespace std in ModifiedGeneralClustering.h)
Initial import of rst docs.
Docstring fixes.
Fix typo in class name.
Remove old testing framework.
Update README.
Update distutils info.
Move examples to a common place.
Remove deprecated path hack.
Enabling test_speed.
Neighbors refactoring, predict method added.
Update examples and move them to a common place.
Update README. Specify dependencies more clearly.
Remove old testing framework.
Remove rendundant file extension in numpy.ctypeslib.load_library().
Fix doctests in Neighbors.
Move (and update) documentation for em module.
Sphinx extensions.
Add ReST docs.
Delete module manifold_learning.regression.neighbors.utilities
Fix functions that depend on Neighbors.
Add AUTHORS file
Update web page.
Change location of header files.
Fix typos in setup.py.
Add needed files from numpy.testing.
Remove genetic algorithms.
Remove references to deleted classes.
Cosmetic cleanup of setup.py.
Use KDTree from scipy.spatial in Nearest Neighbors.
Add tests for Nearest Neighbor Algorithm.
Update KNN examples.
Add a MANIFEST.in file.
Remove C header file for neighbors.
Fix relative imports in manifold_learning/regression/tests.
Fix bug in barycenters.
Fix imports of scikits.optimization.
API changes for Neighbors broke some tests.
Fix bug in knn example.
Deleted MANIFEST.in
Fix typo in web page.
Clean some setup.py files.
Deleted common/myfunctools.py.
Refactor test_neighbors.
0.1 beta release
Gael Varoquaux (1):
MISC: Cosmetic changes in the rst formatting of the README
-----------------------------------------------------------------------
No new revisions were added by this update.
--
Alioth's /usr/local/bin/git-commit-notice on /srv/git.debian.org/git/debian-science/packages/scikit-learn.git
More information about the debian-science-commits
mailing list