Thoughts/comments from May 30

[mgelbart]

• Make it easier to run with various options, without having to copy/paste from run.sh and without having to comment out lines (use command line arguments to specify everything)
• To implement the "real world" problems, you can use this. The parameter alpha corresponds to either lambda or lambda/2. The parameter gamma probably corresponds to 1/(2sigma^2); see here.

You can also use sklearn for the cross-validation.

Notes for BayesOpt: if you use Python 2, change "4" to "4.0" etc wherever you're dividing two integers.