I took a stab at addressing MADLIB-1118 (IC is taking too long for elastic
net). I'm not sure if the approach (updating elastic_net_train tolerance
value (1e-6 --> 1e-2) is appropriate. Although it does deliver the expected
reduction in execution. As the jira suggested, I tried to reduce the data
set and found the execution time was not significantly reduced.
Additionally, I added a comment to the PR which showed how I approached
tweaking the test suite.
If you have a chance, would you mind reviewing the PR?