JPMorgan Chase and the Amazon Advanced Solutions Lab have announced a coordinated research effort to explore quantum computing techniques for complex optimization tasks encountered in finance.
The collaboration produced three technical papers that outline methods for applying near-term analog neutral-atom quantum machines - specifically Rydberg hardware - to problems that are traditionally handled by classical solvers. The teams designed approaches that allow quantum processors to function as co-processors alongside conventional optimization algorithms.
Portfolio optimization and rebalancing
One paper focuses on portfolio construction and rebalancing. The researchers describe a decomposition pipeline that, according to the study, can shrink the size of realistic portfolio optimization instances by roughly 80% while preserving solution quality. In tests on large problems with up to 1,500 variables, the pipeline achieved an approximate 3x improvement in time-to-solution.
Quantum compilation for graph problems
A second paper introduces a quantum compilation toolkit tailored to the maximum independent set problem on Rydberg atom arrays. The toolkit markedly lowered qubit requirements for real-world graph instances. An illustrative example provided compares the Cora citation graph, which has about 2,700 nodes: prior approaches would have needed on the order of 29 million qubits, while the new compilation method reduced that requirement to the scale of tens of qubits for practical execution.
qReduMIS: hybrid quantum-classical co-processing
The third paper presents qReduMIS, a hybrid algorithm that pairs exact polynomial-time reduction techniques with measurement data from quantum runs. In this hybrid model, quantum devices act as co-processors to support classical reduction steps. Experiments carried out on QuEra’s Aquila device accessed via Amazon Braket produced average success rates above 89% on hard problem instances cited in the research.
The empirical work included runs using up to 231 qubits on QuEra’s Aquila machine. All three tools were developed with analog neutral-atom Rydberg systems in mind and are intended to bridge current classical methods with near-term quantum hardware capabilities.
Implications and context
The papers collectively map out practical techniques for reducing problem dimensions, lowering hardware requirements, and integrating quantum measurement data into classical reduction logic. The research emphasizes co-processing workflows that leverage Rydberg-based hardware available through cloud access points such as Amazon Braket.