Machine Learning-Guided Compilation for Reconfigurable Logical Quantum Architectures

ORAL

Abstract

Early fault-tolerant quantum computers will face a fundamental resource imbalance: limited physical qubits but a demand to execute increasingly large logical circuits. To make efficient use of these constrained resources, future processors will likely adopt densely packed and reconfigurable logical layouts. These conditions challenge conventional static mapping approaches that assume fixed logical qubit placements. We present a machine learning–guided Monte Carlo Tree Search (MCTS) compiler that adapts to these settings by exploring placement and routing decisions dynamically. Beyond improving compilation, this framework provides a means to explore how dense, reconfigurable logical architectures can best exploit the limited resources of early fault-tolerant quantum processors.

*This work was supported by the U.S. Department of Energy, Office of Science, Office of Advanced Scientific Computing Research through the Accelerated Research in QuantumComputing Program MACH-Q Project. This research used resources of the National Energy Research Scientific Computing Center (NERSC), a Department of Energy Office of ScienceUser Facility using NERSC award DDR-ERCAPm4141. This work was also funded by the National Science Foundation (NSF) through the Challenge Institute for Quantum Computation (CIQC).

–

Presenters

  • Mathias Weiden

    • University of California, Berkeley
    • UC Berkeley

Authors

  • Mathias Weiden

    • University of California, Berkeley
    • UC Berkeley

  • Justin Kalloor

    • UC Berkeley

  • John D Kubiatowicz

    • University of California, Berkeley
    • UC Berkeley

  • Costin C Iancu

    • Lawrence Berkeley National Laboratory