• 7 November 2025
• 9:00 EST

Note that this seminar will take place at 9am Eastern Time / 3 pm CET Abstract: In recent years, CubeSats, a class of small satellites consisting of several 10-cm cubic units, have been increasingly employed in astronomical missions. In X-ray astronomy, celestial sources must be observed from space to avoid attenuation of X-ray photons by the Earth's atmosphere. As a result, the advantages of CubeSats (such as their low cost and short development times) are gaining wider recognition, leading to an increasing number of CubeSat missions being planned. NinjaSat is the first Japanese 6U CubeSat observatory designed to observe bright X-ray sources, such as black holes and neutron stars. It was launched on November 11, 2023, by the SpaceX Transporter-9 mission and observed 32 X-ray sources during its two-year operation. We successfully demonstrated that CubeSat observations can provide valuable astronomical X-ray data, highlighting NinjaSat's pioneering role as a compact yet uniquely powerful observatory in time-domain astronomy. In this talk, I present an overview of NinjaSat, including the development of the gas X-ray detector, which is one of the most suitable choices for CubeSat applications, and student-led in-orbit operations. In addition, I present observational results of X-ray bursters, which are thought to be key sites for the production of proton-rich heavy nuclei.

• 7 November 2025
• 3:00 EST

Particle accelerators continue to grow in complexity, requiring tighter tolerances and better tuning methods. Light sources are an especially demanding case due to strict requirements on beam uptime and stability. Recently, machine learning (ML) methods like Bayesian optimization and reinforcement learning have shown promise for experimentally finding and maintaining optimum states. However, a lot of preparation and development is required to use ML tools effectively. In this talk, we will review the commissioning of the Advanced Photon Source (APS) facility, which has just completed an upgrade to become one of the world's brightest storage-ring light sources. We will then dive into details of ML use-cases such as first stored beam search, lattice optics inference, and multi-objective (MO) tuning of injection efficiency and lifetime. MO tuning is an especially interesting high-dimensionality task that was previously intractable with classic optimization methods. We will highlight novel Bayesian optimization algorithm and methodology improvements, and show results of several experimental optimization runs up to 24 dimensions, benchmarking efficiency as compared to standard genetic algorithms. We will conclude with an overview of ongoing ML applications at other facilities, and a discussion of integration into standard control room toolkits.

• 7 November 2025
• 5:30 EST