Researchers Achieve Speedup with Parallel Quantum Algorithms

The team evaluated three distinct strategies for distributing calculations within the Variational Quantum Eigensolver algorithm using the CUNQA platform. The method enables more efficient computation by dividing quantum tasks between multiple processing units with subsequent classical analysis of the results. These workflows address challenges posed by inconsistencies in noise levels found in current quantum hardware. The team investigated how to split up calculations for the Variational Quantum Eigensolver algorithm using multiple processing units without requiring direct quantum links between them.

This approach is particularly relevant given current limitations in linking quantum processors together effectively. They tested three methods, distributing individual calculation ‘shots’, parts of the overall computational circuits or entire candidate solutions, utilising a software platform that simulates realistic quantum hardware with varying error rates. Researchers are exploring ways to use the power of multiple early prototype quantum computers, akin to dealing with an old television signal plagued by static, to solve complex problems more efficiently.

The team evaluated strategies for distributing calculations within the Variational Quantum Eigensolver algorithm, essentially a recipe for finding the lowest energy state of molecules or materials using both quantum processors and conventional computers. This approach is particularly useful because current limitations prevent effective linking between these quantum processing units; instead, they investigated dividing up large computational tasks into smaller ones that can be handled independently, similar to several people each painting a section of a very long fence.

These workflows address challenges arising from inconsistencies in error rates across different quantum devices. Varying noise profiles impact optimisation trajectories and overall accuracy when splitting computations in this way.

Variational Quantum Eigensolver speedup via distributed processing on simulated noisy quantum hardware

A factor of five improvement in speedup was achieved by distributing calculations within the Variational Quantum Eigensolver algorithm compared to sequential processing when utilising six virtual quantum processing units; previously, this level of parallelisation proved unattainable due to limitations in classical communication bandwidth between processors. Researchers at Galicia Supercomputing Centre and Universidade de Santiago de Compostela evaluated three distinct approaches, shot-level, circuit-level, and population-level parallelism, using their CUNQA platform which simulates realistic noisy hardware environments. This work addresses challenges posed by inconsistencies in error rates across different quantum devices, enabling more efficient computation through division of tasks with subsequent classical analysis of results.

Researchers at Galicia Supercomputing Centre and Universidade de Santiago de Compostela detailed the performance gains achieved through parallelisation of the Variational Quantum Eigensolver algorithm; specifically, executing each quantum circuit measurement on separate virtual processors, shot-level distribution, reduced overall computation time by a factor of five compared to sequential processing utilising six units. Similar efficiency improvements were demonstrated when distributing workload across the CUNQA platform’s simulated noisy hardware via circuit-level parallelism which focuses on concurrent calculation of gradients and observable values. Employing gradient-free optimisers like Differential Evolution also benefited from population-level approaches allowing for simultaneous evaluation of multiple candidate solutions.

Mitigating variable error rates through distributed processing unlocks potential gains with nascent

The pursuit of more powerful computation drives exploration into harnessing multiple early prototype quantum computers; however, adding processing units does not guarantee success if each unit behaves differently due to inherent inconsistencies in error rates. Distributing calculations can accelerate results but relies on classical post-processing to reconcile disparate outputs, a technique that may not fully compensate for unpredictable noise profiles found in real hardware. The team evaluated different methods of splitting complex tasks, such as finding the lowest energy state of a molecule, across these simulated units acknowledging inconsistent error rates present a significant challenge to achieving speedup and demonstrated how it improves computational efficiency. This approach addresses limitations posed by imperfections in current prototype quantum computer’s error rates allowing exploration of parallelisation strategies even with imperfect hardware. They employed an emulation platform called CUNQA to evaluate impacts on both efficiency and accuracy when dealing with noisy devices while testing their method using a system size of six spins for the Transverse Field Ising Model. Consequently, distributing calculations within the Variational Quantum Eigensolver, a technique used to find the lowest energy state of molecules and materials, can sharply improve computational speed without requiring direct connections between virtual processing units.

Distributing computations improved performance of the Variational Quantum Eigensolver by enabling concurrent calculation of gradients and observable values via circuit-level parallelism. These findings suggest that calculations can be accelerated without needing direct links between individual processing units when utilising this distributed method.