Eric Liu

Nope, I haven't had a chance to dive deeper into this. A team at Cornell where some of the original research for MOE was done put a paper up on...

If you have some results demonstrating the value/trade-off, I would love to see them :)

This would be better solved by something like #274. Having gigantic argument lists where half of them can be null to engage some default behavior is worse than naming these...

I'm downgrading this to P4. Someone could do it if they're interested in learning but it's not important. Reasoning: CPU EI code is fully ping tested. #219 will run the...

This is already done for the constructors of NewtonParameters and GradientDescentParameters through #138.

when this happens, we should edit `conda-moe/README.md` to point to the official channels

It'd also be nice to have a decorator for timing functions. Something like what's here: http://stackoverflow.com/questions/5478351/python-time-measure-function although due to this: http://stackoverflow.com/questions/7680446/get-python-functions-owning-class-from-decorator we'll prob want to accept some extra descriptor arguments...

note: a simple timing context manager was added through the commit for #82 and I added timing calls to the most expensive parts of each REST endpoint. I'm leaving this...

I think option 2 is preferred for now. the current views/schemas/ folder should also move to the clientlib.

tolerances decreased so these tests pass. I'm bumping the priority of this ticket down. It is now about making these tests more repeatable (i.e., by setting RNG)