2025-07-23 –, Main Room 6
We supercharge quantum-repeater simulations by presenting Julia packages that incorporate a recent technique from machine learning, stochastic automatic differentiation. To demonstrate its usefulness, we optimize rate-fidelity trade-offs and optimally place repeaters in a 2D plane. We observe spontaneous symmetry breaking and discover that the required number of repeaters scales only with the square root of the network area.
Future quantum networks are envisioned to create entanglement between end nodes using intermediate nodes called quantum repeaters. However, before useful real-world quantum repeaters can be constructed, major breakthroughs are required both in the engineering of quantum devices and in understanding how to configure those devices. Over the last few years, Monte Carlo simulations have rapidly become one of the most important tools used to study quantum networks. They can be used to determine, given a specific hardware quality and device configuration, how good a network would be. However, they struggle to provide meaningful information about how hardware should be improved and how devices should be configured to build the best quantum networks as soon as possible.
In this talk, which is based on our paper titled “Optimization of Quantum-Repeater Networks using Stochastic Automatic Differentiation”, we discuss how to overcome this problem using a recent technique from machine learning called stochastic automatic differentiation (stochastic AD). Stochastic AD provides a framework for automatic differentiation that is compatible with discrete random programs, allowing the extraction of derivative estimates that are both accurate and precise, something that is notoriously difficult in the presence of discrete randomness. As quantum networks are rife with discrete randomness, their simulations are discrete random programs. Therefore, Stochastic AD can be used to extract derivative information that directly reveals the most crucial hardware parameters and can be used to optimize networks using gradient-based optimization techniques.
We have created an open-source Julia package (QuantumNetworkRecipes.jl) that implements Monte Carlo simulations of quantum repeaters in a way that is compatible with StochasticAD.jl, the first and currently only implementation of stochastic AD. We use this package to provide concrete examples of how useful a repeater simulation supercharged with stochastic AD can be, such as by efficiently tuning rate-fidelity trade-offs in a repeater chain through gradient descent and determining the most crucial parameters in a quantum network. As our most significant example we address the optimal placement of quantum repeaters in a 2D plane. We have developed an open-source Julia package (RepeaterPlacement.jl) to find optimal repeater locations using gradients obtained from QuantumNetworkRecipes.jl and StochasticAD.jl. This has allowed us to reveal an interesting case of spontaneous symmetry breaking and discover that, in the studied case, the required number of repeaters only scales with the square root of the total area of the network. We believe these examples demonstrate that integration of stochastic AD in quantum-network simulators could accelerate the research and development of functional quantum networks.
Guus Avis is a postdoctoral researcher specialized in the analysis and design of quantum networks. One of his interests is investigating how real-world imperfect quantum hardware can best enable useful network protocols, which requires a combination of device physics and protocol design.
After obtaining both a Bachelor of Science in physics and a Bachelor of Arts in philosophy at the University of Groningen, Guus went on to obtain a Masters in Science in theoretical physics from the University of Utrecht, where he specialized in cosmology. For his doctoral studies, Guus switched to the exciting field of quantum technology and joined the Delft University of Technology. There he obtained his PhD on quantum networks under the guidance of Stephanie Wehner, working as a part of both QuTech and the Quantum Internet Alliance.
After graduating, Guus has left The Netherlands to experience what it is like to live above sea level. At UMass Amherst, under the wing of Stefan Krastanov, Guus continues to contribute towards making quantum networks a reality.