Eric Liu
Eric Liu
RELATED TO: GH-191 (deleting rows from cholesky) It's probably easiest to prototype this in Python and then implement in C++. This would make computing covariance matrices (and cholesky factors) from...
RELATED TO: #314 In #257, I catch cholesky failures and use a combination of the SVD and QR factorizations to handle numerically SPSD matrices smoothly. This is probably overkill but...
`OptimizableGpPrettyView.get_params_from_request()` overrides the default `GpPrettyView.get_params_from_request()` behavior that just deserializes via colander. The deserialization of optimization parameters is more complex b/c some child nodes of OptimizationInfo are dependent on other nodes...
At the moment, e.g., `OptimizableGpPrettyView.get_params_from_request` is a hacky mess that bypasses colander almost entirely. It reads user input from raw json, pulls a default value, and then overwrites defaults with...
As of #214, Python (REST interface, via colander) simply checks that all hyperparameters are positive floats. This is true for current covariance functions, but it need not always be true....
#82 fixed a bug where users could specify INT_MAX threads and crash the server. It introduced `MAX_ALLOWED_NUM_THREADS` as a constant. It'd be nice if server admins could set this value...
On the full-install page, we should have a little table of MOE specific cmake variables (like specifying python, C compiler, etc). Currently these guys are kind of scattered throughout the...
Right now we have some `/examples` that access `moe/easy_interface` modules. These modules in turn require some _basic_ components of MOE (constants.py and data_containers.py). Unfortunately, at the moment users have to...
RELATED TO: #279 In simple_endpoint.py, we provide a few wrappers around MOE REST endpoints. At the moment, they're pretty poorly documented. With 3 (at the time of this writing) functions...
I see things like: ``` historical_data.json_payload() covariance.get_json_serializable_info() domain.get_json_serializable_info() _build_gp_historical_info _build_gp_historical_info_from_points_sampled ``` There's no reason for these to all be named differently when they all do the same task on their...