Working with sparse parameterizations in optimizer

The optimizer framework can be used to reduce overhead significantly when solving many similiar problems, by pre-compiling a model parameterized in set of parameters which can change.

Unfortunately, in some cases the precompiled approach can become slower than a standard modelling approach. The mistake is typically a parameterization which does not exploit sparsity. Learn how to avoid this…