Constrained Optimization with HGDL
Here we show how constrained optimization can be performed with the HGDL optimization package. This optimizer is available in fvGP and gpCAM and is activated automatically when the user initiates the asynchronous training. The great strength of this optimizer is that it can run in parallel and distributed over many nodes, uses local and global optimizers in tandem, can remove found optima, and runs asynchronously. Good luck :)
Here we show how constrained optimization can be performed with the HGDL optimization package. This optimizer is available in fvGP and gpCAM and is activated automatically when the user initiates the asynchronous training. The great strength of this optimizer is that it can run in parallel and distributed over many nodes, uses local and global optimizers in tandem, can remove found optima, and runs asynchronously. Good luck :)
This is a new feature and is working with HGDL version 2.1.0.
The image shows the result of the run. We see, in green, the constraints. The found solutions are in red.