| commit | 95e8496aceed3692f9988ae8550c35701b731ee8 | |
| author | eibe <eibe@e0a1b77d-ad91-4216-81b1-defd5f83fa92> | |
| Wed, 9 Dec 2009 00:56:39 +0000 (9 00:56 +0000) | ||
| committer | eibe <eibe@e0a1b77d-ad91-4216-81b1-defd5f83fa92> | |
| Wed, 9 Dec 2009 00:56:39 +0000 (9 00:56 +0000) | ||
| tree | 7d2333ae8e43131c5f26a76bafa32e7cd1c887ca | treesnapshot (tar.gz zip) |
| parent | 20f4d110f44f32b7d3577f47add0823bc8b340bc | commitdiff |
New version that stops boosting when exponential loss on training data does not decrease further. Shrinkage should now also work better because predictions are shrunk towards a ZeroR model.
git-svn-id: https://svn.scms.waikato.ac.nz/svn/weka/trunk/weka@6136 e0a1b77d-ad91-4216-81b1-defd5f83fa92
git-svn-id: https://svn.scms.waikato.ac.nz/svn/weka/trunk/weka@6136 e0a1b77d-ad91-4216-81b1-defd5f83fa92
| src/main/java/weka/classifiers/meta/RealAdaBoost.java | diffblobblamehistory |