Evolutionary Circuit Optimization Reduces Quantum Communication Costs by 89% for Distributed Computing

Distributed quantum computing promises to overcome limitations of current quantum processors, but efficiently allocating tasks and minimising communication between processing units presents a significant challenge. Leo Sünkel, Jonas Stein, and Gerhard Stenzel, from the Institute for Informatics LMU Munich, alongside colleagues, now demonstrate a novel approach to circuit optimisation inspired by evolutionary algorithms. Their method tackles this communication bottleneck by intelligently restructuring quantum circuits to reduce the number of necessary connections between quantum processing units. Results show this technique dramatically cuts the demand for global gates, by over 89% in some cases, while maintaining high accuracy and successful problem-solving, and also reduces communication cost by up to 19% when tailored to a specific network layout, representing a substantial step towards practical, scalable quantum computation.

In this work, researchers evaluate an evolutionary algorithm (EA) to optimise circuits for execution within the Distributed Quantum Computing (DQC) paradigm, with a focus on reducing required communication. The approach is evaluated using a state preparation task implemented with Grover circuits, demonstrating a reduction of more than 89% in the number of required global gates while maintaining high fidelity and accurate solution extraction. Furthermore, the method successfully reduces both circuit depth and the number of CX gates employed. Experiments also explore circuit optimisation tailored to a specific network topology, assigning each qubit before the optimisation process begins.

Genetic Algorithms Optimise Quantum Teleportation Costs

This research details an investigation into optimizing the cost of quantum teleportation in distributed quantum computing (DQC) using evolutionary algorithms, specifically genetic algorithms. The core challenge addressed is efficiently partitioning a quantum circuit and mapping it onto a network of quantum processing units (QPUs) to minimize the communication overhead, or teleportation cost, between them. The team aimed to balance computational load and communication demands to achieve optimal performance. The researchers employed a genetic algorithm to explore potential circuit partitions and QPU assignments.

Each solution represents a specific way of dividing the circuit’s gates and qubits across different QPUs. The algorithm evaluates each solution using a fitness function that quantifies the total teleportation cost; lower costs indicate better solutions. Standard genetic operators, including selection, crossover, and mutation, are used to evolve a population of solutions over time. The research considered various network topologies connecting the QPUs. To refine the partitioning process, the team incorporated the Kernighan-Lin algorithm as a local search operator within the genetic algorithm.

Results demonstrated that the genetic algorithm consistently finds solutions with significantly lower teleportation costs, proving its effectiveness in optimizing circuit partitioning. The optimal partitioning strategy depends on the network topology, highlighting the importance of considering network structure during optimization. This research presents a novel application of genetic algorithms to the problem of optimizing teleportation cost in DQC. The proposed strategy effectively minimizes communication overhead, and the findings provide insights into the impact of network topology on optimal partitioning.

These results have practical implications for the design and implementation of DQC systems. Future research will explore more sophisticated fitness functions, hybrid algorithms combining genetic algorithms with other optimization techniques, and evaluation on more realistic network models. The team also plans to develop automated tools for circuit partitioning and integrate this strategy into quantum compilers.

Quantum Circuit Optimisation via Evolutionary Algorithm

This research demonstrates a novel evolutionary algorithm capable of significantly optimising quantum circuits for execution in distributed computing environments. Experiments utilising Grover circuits reveal substantial reductions in circuit complexity. The algorithm achieved an 89% reduction in required global gates while maintaining high fidelity and accurate solution extraction. Further analysis focused on minimizing circuit depth, yielding reductions of 36%, 32%, and 27% for circuits with 4, 5, and 6 qubits respectively. The team also successfully minimized the number of CX gates, achieving rates between 39% and 83% depending on the circuit size.

A key finding is the impact of incorporating fidelity weighting into the optimization process. Circuits optimized solely for metric reduction exhibited altered state distributions, potentially impacting performance under noisy conditions. Experiments designed to reduce communication cost in networked quantum processing units (QPUs) delivered promising results. A 6-qubit circuit optimized for a 3-node network achieved a 14. 33% reduction in communication, while an 8-qubit circuit on a 4-node grid demonstrated a 19.

575% reduction in the number of hops between QPUs. The team measured mean fidelities ranging from 0. 966656 to 0. 972087 across various optimization strategies, confirming that the optimized circuits consistently prepared states capable of yielding the correct solution. These results demonstrate the potential of evolutionary algorithms to address the challenges of quantum circuit optimization, paving the way for more efficient and scalable distributed quantum computation. While the optimized circuits may not be functionally equivalent to the originals, they consistently achieve the desired output, offering a valuable tool for future quantum compilation stacks.

Evolutionary Optimisation Cuts Quantum Communication Costs

This work demonstrates the effectiveness of an evolutionary algorithm in optimising quantum circuits for execution in a distributed quantum computing paradigm. The algorithm successfully reduces communication overhead, achieving up to a 30% reduction in required non-local gates while maintaining high fidelity, above 0. 9 in all instances, and ensuring the circuits continue to produce correct solutions. Furthermore, the approach reduces communication cost by up to 19% when optimising circuits for specific qubit assignments within a network. The authors acknowledge that further improvements may be possible through extended algorithm execution or more comprehensive hyperparameter optimisation. Future research directions include integrating this optimisation approach into a complete compilation stack for distributed quantum computing and adapting it for a wider range of applications requiring functionally equivalent circuits. While the results demonstrate significant potential, the degree of optimisation required depends on the specific use case, with even modest reductions proving valuable in certain contexts.