Eric Liu
Eric Liu
C++'s gradient descent code currently does not support any kind of averaging. We should implement polyak-ruppert averaging. This is already done in `moe.optimal_learning.python.python_version.optimization.GradientDescentDescentOptimizer.optimize` so porting it should be straightfoward. This...
Getting the same question in #380 as #369 means re-emphasizes our need for: 1. a wiki 2. updates to our existing FAQ (with questions like these and others we've received...
BLOCKED BY: #378 After #378 makes it so that you can select covariance types through the Python interface, we should make this feature accessible through the REST interface. 1. set...
Right now, the boost C++/Python interface assumes the user wants `SquareExponential`. Operations that require covariance classes currently just accept a vector of hyperparameters, and then they always construct the sqexp...
Right now, the Python optimal_learning library uses only numpy arrays internally. But the interface to the C++ library only accepts python lists. The usual process to transfer data in is:...
When we first made the move to autodoc the C++ with doxygen+sphinx+breathe, I essentially just put double backticks around all existing mathy expressions so that at least the expression would...
I often wrote things like: ``` explanation: ``some math`` more explanation ``more math`` ``` after doc-ifying, the carriage returns are removed and this gets blobbed into one big messy paragraph....
Right now, if we have: ``` in .hpp: /*!\rst blah blah \param :blah: blah \return stuff \endrst*/ void foo(); in .cpp: /*!\rst implementation details \endrst*/ void foo() {} ``` the...
After #353, the C++ autodoc pipeline (doxygen, sphinx, breathe) now captures the source files in the `cpp/gpu` directory but it does so by flattening out the gpu directory. Plus, the...
`_compute_expected_improvement_qd_analytic` and `_compute_expected_improvement_1d_analytic` both finish by calling `numpy.fmax`. If either argument to fmax is NaN, it returns the non-NaN arg: http://docs.scipy.org/doc/numpy/reference/generated/numpy.fmax.html#numpy.fmax numpy.maximum propagates nans; we may want that instead. I'm...