Dorian Daimer

Dorian Daimer

PhD Student at UH Manoa, working on physical limits of information processing

Publications

Thermodynamically rational decision making under uncertainty

Published in arXiv preprint arXiv:2309.10476, 2023

We use the framework of partially observable information engines to obtain an analytical characterization of thermodynamically rational agent behaviour for a simple, yet non–trivial example of a Maxwells demon operating under uncertainty. Our results provide the first complete detailed physical understanding of a decision problem under uncertainty. Download paper here

Recommended citation: Daimer, D. & Still, S. (2023). Thermodynamically rational decision making under uncertainty. arXiv preprint arXiv:2309.10476.

The physical observer in a Szilard engine with uncertainty

Published in arXiv preprint arXiv:2309.10580, 2023

We use the fact that an algorithm for computing optimal strategies can be directly derived from maximizing overall engine work output in generalized partially observable information engines. For a stylizedly simple decision problem, we discover interesting optimal strategies that differ notably from naive coarse graining. They inspire a model class of simple, yet compelling, parameterized soft partitionings. We analyze and compare optimal strategies for three different observer classes: (1) optimal observers, (2) observers limited to the parameterized soft partitionings introduced here and (3) observers limited to coarse graining. While coarse graining based observers are outperformed by the other two types of observers, there is no difference in performance between unconstrained, optimal observers and those limited to soft partitionings. The parameterized soft partitioning strategies allow us to compute key quantities of the decision problem analytically. Download paper here

Recommended citation: Daimer, D. & Still, S. (2023). The physical observer in a Szilard engine with uncertainty. arXiv preprint arXiv:2309.10580.

Partially Observable Szilard Engines

Published in New J. Phys. 24 073031, 2022

This papers provides the first example of a Partially Observable Szilard Engine. The partially observable engine is obtained from a standard Szilard Engine by inserting the divider at an angle instead of horizontally. The resulting family of engines (parameterized by the angle) is analyzed in great detail within the framework of Generalized Partially Observable Information Engines. Interestingly, optimal data encodings for the partially observable engines are probabilistic in general and we provide a scheme to construct these probabilistic data representations in a simple way. Download paper here

Recommended citation: Still, S., & Daimer, D. (2022). Partially Observable Szilard Engines. New J. Phys. 24 073031.