Using Machine Learning to Design Quantum Network Tomography Protocols

Quantum networks are interconnected quantum computers that exchange quantum information. Realizing such networks is a foundational step toward building a quantum internet, which is essential for meeting the demands of large-scale quantum computing. However, quantum information is highly sensitive to noise, making reliable communication a major challenge. Quantum Network Tomography (QNT) helps by analyzing how quantum information behaves across a quantum network. The Quantum Fisher Information Matrix (QFIM) measures how well a QNT protocol captures error-related data. Designing protocols that maximize the QFIM is a difficult and unsolved problem, where traditional methods are often computationally intensive and struggle with the complexity of real-world quantum networks. Machine learning offers a promising alternative. In this project, we explore a machine learning-based approach to QNT by training a Reinforcement Learning Agent (RLA) to distribute qubits across simulated networks. The RLA discovers strategies that outperform hand-crafted protocols by achieving higher QFIM values. Our results show that machine learning can produce adaptive, efficient, and scalable QNT protocols, highlighting the potential of machine learning-based approaches to QNT.