29 April 2024 in Machine Learning: Science and Technology
The generation of cold atom clouds is a complex process which involves the optimization of noisy data in high dimensional parameter spaces. Optimization can be challenging both in and especially outside of the lab due to lack of time, expertise, or access for lengthy manual optimization. In recent years, it was demonstrated that machine learning offers a solution since it can optimize high dimensional problems quickly, without knowledge of the experiment itself. In this paper we present results showing the benchmarking of nine different optimization techniques and implementations, alongside their ability to optimize a rubidium (Rb) cold atom experiment. The investigations are performed on a 3D 87Rb molasses with 10 and 18 adjustable parameters, respectively, where the atom number obtained by absorption imaging was chosen as the test problem. We further compare the best performing optimizers under different effective noise conditions by reducing the signal-to-noise ratio of the images via adapting the atomic vapor pressure in the 2D+ magneto-optical trap and the detection laser frequency stability.
4 December 2024 in Solar RRL
The significance of color aesthetics in photovoltaic (PV) modules gains importance, especially in design‐centric applications like building‐integrated PVs. Color filters based on distributed Bragg reflectors, consisting of alternating thin‐film layers of different refractive indices, can modify the appearance of standard silicon modules. This approach is also extended to optimize the color appearance of emerging PV technologies such as perovskite solar cells, which typically exhibit a less appealing gray–brownish appearance. In this contribution, perovskite solar‐cell stacks combined with MorphoColor color filters are presented. Angular‐resolved reflectance simulations based on wave optics and ray tracing with experimental data are validated, and the color appearance from various viewing angles is evaluated. Additionally, the impact of individual layers on color appearance and the maximum achievable short‐circuit current density in the perovskite solar cell is investigated. By applying Bayesian optimization, the color distance is minimized to the targeted appearance. Tailoring the bridging layers between the color filter and the perovskite solar cell is found to strongly influence the color impression due to the coherently combined color filter and perovskite solar cell. The presented color optimization concept allows to customize the aesthetics of emerging PV thin‐film technologies such as perovskite solar cells.
28 November 2025 in Machine Learning: Science and Technology
Bayesian optimization with Gaussian process (GP) surrogates is a popular approach for optimizing expensive-to-evaluate functions in terms of time, energy, or computational resources. Typically, a GP models a scalar objective derived from observed data. However, in many real-world applications, the objective is a combination of multiple outputs from physical experiments or simulations. Converting these multidimensional observations into a single scalar can lead to information loss, slowing convergence and yielding suboptimal results. To address this, we propose to use multi-output GPs to learn the full vector of observations directly, before mapping them to the scalar objective via an inexpensive analytical function. This physics-informed approach retains more information from the underlying physical processes, improving surrogate model accuracy. As a result, the approach accelerates optimization and produces better final designs compared to standard implementations.
10 December 2025 in Physical Review B
19 November 2024 in Quantum Science and Technology
State engineering of quantum objects is a central requirement for precision sensing and quantum computing implementations. When the quantum dynamics can be described by analytical solutions or simple approximation models, optimal state preparation protocols have been theoretically proposed and experimentally realized. For more complex systems such as interacting quantum gases, simplifying assumptions do not apply anymore and the optimization techniques become computationally impractical. Here, we propose Bayesian optimization based on multi-output Gaussian processes to learn the physical properties of a Bose–Einstein condensate within few simulations only. We evaluate its performance on an optimization study case of diabatically transporting the quantum gas while keeping it in its ground state. Within a few hundred executions, we reach a competitive performance to other protocols. While restricting this benchmark to the well known Thomas–Fermi approximation for straightforward comparisons, we expect a similar performance when employing more complex theoretical models, which would be computationally more challenging, rendering standard optimal control theory protocols impractical. This paves the way for efficient state engineering of complex quantum systems including mixtures of interacting gases or cold molecules.