search

Yelp / MOE

A global, black box optimization engine for real world metric optimization.
Other
1.31k stars 139 forks source link

Sclark 384 blog ab example #403

Closed sc932 closed 9 years ago

sc932 commented 9 years ago

***** PEOPLE ***** Primary reviewer: @suntzu86

Reviewers:

***** DESCRIPTION ** Branch Name: sclark_384_blog_ab_example Ticket(s)/Issue(s): Closes #384

***** TESTING DONE ***** manually run make test make style-test

suntzu86 commented 9 years ago

i dont think this file passes pep257 and our other style checks (e.g., describing fcn inputs). is that a concern? yes it's an example but probably/hopefully one that will be read a lot

suntzu86 commented 9 years ago

high level comments:

  1. input/output descriptors for docstrings
  2. i don't totally follow the data flow through the example. we seem to be calling the objective function in many places for example--where does the simulated experiment round happen?
  3. following 2, it could be nice to have the fcn names be more specific as to which ones are true_objective (should only called to get the parameters for doing 1M bernoulli draws and nowhere else) and which ones are measured_objective

I'm definitely a little confused on the data flow