Observable partitioning under uncertainty in information engines with physically implemented observer memories
Published:
This poster presentation at the APS March Meeting 2023 explores a simple model class of information engines with uncertainty: Observations either carry no uncertainty or maximum uncertainty about the relevant quantity. Within the framework of generalized, partially observable information engines the optimal data representation strategies are found and a concrete physical model to realize these encodings is presented. Two different approximations of the optimal encodings, coarse graining and soft partitioning, are analyzed for comparison.