The computational burden of optimization and memory issues of optimization for large dimensions merely comes from storing the history of the input vector and losses. At the moment this is done by default.
This results in out-of-memory errors when optimizing large systems (>100D).
Solution
Implement a way to overwrite the current decision vector instead of appending to f3dasm.ExperimentData.
Problem
The computational burden of optimization and memory issues of optimization for large dimensions merely comes from storing the history of the input vector and losses. At the moment this is done by default.
This results in out-of-memory errors when optimizing large systems (>100D).
Solution
Implement a way to overwrite the current decision vector instead of appending to
f3dasm.ExperimentData
.