Our JCSS paper (“Robust optimization in the presence of uncertainty: A generic approach”) is out - a nice birthday present btw!
The highlights of the paper are:
- A new approach to robust optimization in presence of uncertainty ROPU is proposed.
- ROPU takes two typical instances and doesn’t assume any special noise model.
- ROPU measures task specific similarity of instances, i.e., its input relevance.
- The similarity measure detects if given instances are not similar or too noisy.
- Instance similarity favors good localization (!) of solutions rather than costs.