Adam Śmiechowski
Adam Śmiechowski
Looking through Greg's changes
 Change those default values to false to avoid unexpected behaviors.
A Gaussian Blobl is used in test_optimization_with_context right now but it's an unstable test and it's unclear why. Pulling Gaussian Blob as a separate objective function would allow us to...
Right now we support only two DOE schemes: * Guided (by maximizing utility function over surrogate model predictions) * Random However, there is a plethora of other DOE schemes we...
This will be a staged process: - [] Allow for specifying objective spaces with more than one dimensions. (specify which of the objectives to optimize) - [] Allow for registering...
It's really simple to implement (given the presence of both Random Search and Glow Worm optimizers) but it should give us a substantial improvement over Random Search.
We should get a substantial boost in performance if we remember good positions from previous calls. The intuition is that the positions of optima don't necessarily change that much between...
Haven't debugged it in too much detail yet, but it appears that when we send the data over gRPC some reordering takes place and we fail to account for it....