Frank McQuillan created MADLIB-1384:
---------------------------------------
Summary: Change default num_components for SVM to 100
Key: MADLIB-1384
URL: https://issues.apache.org/jira/browse/MADLIB-1384
Project: Apache MADlib
Issue Type: Improvement
Components: Module: Support Vector Machines
Reporter: Frank McQuillan
Fix For: v1.17
Currently
http://madlib.apache.org/docs/latest/group__grp__svm.html#kernel_params
says
{code}
n_components
Default: 2*num_features. The dimensionality of the transformed feature space. A
larger value lowers the variance of the estimate of the kernel but requires
more memory and takes longer to train.
{code}
but this produces poor decision boundaries for small num_features. I suggest
we change the default to
{code}
n_components
Default: max(100, 2*num_features). The dimensionality of the transformed
feature space. A larger value lowers the variance of the estimate of the kernel
but requires more memory and takes longer to train.
{code}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
