Craig Sanders
Craig Sanders
AEPsych models currently require lower bounds and upper bounds of parameters to be passed into their constructors, but the model itself should not have to worry about the bounds of...
The AEPsych server has a number of command line options, but these aren't documented any where other than the docstrings. To make this info more visible and help users learn...
Our Github action is supposed to perform linting but it's not configured correctly. Our code should be formatted using flake8 and black, and type-checked with mypy, but only the mypy...
The current model class hierarchy is a bit confusing: - We have the [AEPsychMixin](https://github.com/facebookresearch/aepsych/blob/main/aepsych/models/base.py#L112), which should be made into a proper base class because it's being used as one instead...
AEPsych logs events such as messages being received or sent, but there are a number of improvements that could be made to logging to improve its utility: - The server...
GPs can make use of inducing points. Basically, instead of fitting the posterior to every training point, we can instead fit to to a smaller set of inducing points to...
The ax branch implements a ConfigureableMixin class: https://github.com/facebookresearch/aepsych/blob/ax-backend/aepsych/config.py#L360 This class allows any other AEPsych classes that implement from_config to fetch config arguments in an object-oriented way, allowing code re-use between...