@article{FranzWolfPeriyasamyetal., author = {Franz, Maja and Wolf, Lucas and Periyasamy, Maniraman and Ufrecht, Christian and Scherer, Daniel D. and Plinge, Axel and Mutschler, Christopher and Mauerer, Wolfgang}, title = {Uncovering Instabilities in Variational-Quantum Deep Q-Networks}, series = {Journal of the Franklin Institute}, journal = {Journal of the Franklin Institute}, edition = {In Press, Corrected Proof}, publisher = {Elsevier}, issn = {0016-0032}, doi = {10.1016/j.jfranklin.2022.08.021}, abstract = {Deep Reinforcement Learning (RL) has considerably advanced over the past decade. At the same time, state-of-the-art RL algorithms require a large computational budget in terms of training time to converge. Recent work has started to approach this problem through the lens of quantum computing, which promises theoretical speed-ups for several traditionally hard tasks. In this work, we examine a class of hybrid quantumclassical RL algorithms that we collectively refer to as variational quantum deep Q-networks (VQ-DQN). We show that VQ-DQN approaches are subject to instabilities that cause the learned policy to diverge, study the extent to which this afflicts reproduciblity of established results based on classical simulation, and perform systematic experiments to identify potential explanations for the observed instabilities. Additionally, and in contrast to most existing work on quantum reinforcement learning, we execute RL algorithms on an actual quantum processing unit (an IBM Quantum Device) and investigate differences in behaviour between simulated and physical quantum systems that suffer from implementation deficiencies. Our experiments show that, contrary to opposite claims in the literature, it cannot be conclusively decided if known quantum approaches, even if simulated without physical imperfections, can provide an advantage as compared to classical approaches. Finally, we provide a robust, universal and well-tested implementation of VQ-DQN as a reproducible testbed for future experiments.}, language = {en} } @article{SchloplocherEttengruberSteffens, author = {Schloplocher, Paul Dragos and Ettengruber, Stefan and Steffens, Oliver}, title = {Improvements for building-performance simulations by a comparative finite-element method analysis}, series = {Energy and Buildings}, volume = {278}, journal = {Energy and Buildings}, publisher = {Elsevier}, doi = {10.1016/j.enbuild.2022.112563}, pages = {1 -- 12}, abstract = {This paper presents a method to improve building-performance simulations (BPS) by the comparison and analysis of different approaches based on finite-element method (FEM) models. The lumped parameter method (LPM) is used in several BPS programs and tools. It has the advantage of fast computing times and comparably good accuracy for thermal and energy loads. With the help of detailed FEM simulations, it is possible to further improve the degree of detail and accuracy while maintaining the high simulation speed. In this work, we compare time-dependent results for local temperatures in a generic reference room within a given periode of time. In a second step, the differences between the models with respect to various physical effects are analyzed and used to introduce additional equations into the LPM model in order to improve its accuracy. Thus, we discuss potentials for improvement for BPS and demonstrate a method of a practical implementation. The results show minor differences of less than 0.1 K for radiation and heat transfer, so their level of detail in BPS is appropriate. In these terms, no improvements were pursued within the work. However, the FEM simulation is capable of calculating the internal convective heat transfer and thermal bridges more accurately due to the use of computational fluid dynamics (CFD) and the geometrically precise representation of the FEM model. Here, deviations of up to 1 K in room temperature (convective heat transfer) and up to 0.5 K in wall temperatures (thermal bridges) were pointed out. By improving the LPM with equations obtained from the FEM, these deviations can be reduced to less than 0.2 K, which is a considerable improvement in accuracy.}, language = {en} }