README.md

How to install

• Clone the repo
• cd to the repo folder and run pip install . from the cloned repo under your virtual environment

Examples

Run main.py, the parameters for quantum ansatz is included in test.json.

Below is an example of the test.json configuration file with explanations for each parameter block:

Basic Configuration

{
    "name": "GMVPOpt",
    "parameters": {
        "save": true,
        "seed": 42,
        ...
    }
}

• name: Name of the simulation
• save: Whether to save the simulation results
• seed: Random seed for reproducible results

Encoding Parameters

"enco": {
    "n": 5,
    "N": 15
}

• n: The number of assets in the portfolio
• N: The number of qubits (must satisfy: N/n is an integer, representing the binary block length)

Problem Instance Configuration

"object": {
    "name": "gmvp",
    "method": "test_sets",
    "test_sets": "rand_assets_short_market_n5N15_slec",
    "rand": false,
    "amplify": 1,
    "verbose": true
}

• name: Type of optimization problem ("gmvp" for Global Minimum Variance Portfolio)
• method: Instance generation method ("test_sets" uses pre-generated market data)
• test_sets: Specific test dataset to use (required when method is "test_sets")
• rand: Whether to use purely random generation instead of market data
• amplify: Amplification factor for the problem coefficients
• verbose: Whether to print detailed instance information

Classical Preprocessing

"bf_qd_run": true,
"run": {
    "runT": 100,
    "runMutiple": true
}

• bf_qd_run: Whether to run brute-force and quadratic programming solutions for comparison
• runT: Number of instances to simulate
• runMutiple: Whether to run multiple instances

Quantum Circuit Configuration

"circuit": {
    "qCircAna": {
        "ini": "ws",
        "cost": "Bubble",
        "mixer": {
            "name": "Bubble",
            "distance": 2
        }
    },
    "layer": {
        "p": 2,
        "CirParams": {
            "rand": true,
            "fix": false,
            "file": false,
            "verbose": true
        }
    },
    "circuit_measure": true
}

Initial State and Operators

• ini: Initial state preparation method
  • "ws": Ranked warm-start
  • "maxbias": Max-biased state
  • "aes": Approximate equal-weighted state
  • "rd": Random-weighted state
• cost: Cost operator architecture (default: "Bubble")
• mixer.name: Mixing operator architecture (default: "Bubble" for qubit excitation， "BubblePauli" for parity mixer(XY-mixer) and XXY mixer)
• mixer.distance: Maximum distance between assets (L=1 for nearest-neighbor mixing)

Circuit Layers and Parameters

• p: Number of QAOA layers (each layer contains one cost and one mixing operator)
• CirParams.rand: Whether to generate circuit parameters randomly
• CirParams.fix: Whether to use fixed parameters
• CirParams.file: Whether to read parameters from the previous simulations for the initial parameters (for layerwised optimization)
• CirParams.verbose: Whether to print parameter details
• circuit_measure: Whether to perform post-optimization measurements

Optimization Configuration

"optimizer": {
    "method": "Annealing",
    "optTar": "mean",
    "numIter": 16,
    "numOuterIter": 2000
}

• method: Optimization algorithm ("Annealing" for dual annealing, "COBYLA" for COBYLA)
• optTar: Optimization target ("mean" optimizes expectation values)
• numIter: Number of shots for estimating expectation values
• numOuterIter: Maximum number of expectation value evaluations allowed

Post-processing Configuration

"postprocess": {
    "numMeaus": 2000,
    "postMeasMeth": "measuredict"
},
"consfilter": false

• numMeaus: Number of measurement shots for final post-optimization evaluation
• postMeasMeth: Final measurement method ("measuredict" outputs shot measurement dictionary)
• consfilter: Whether to apply capital constraint filtering during optimization

Simulator Configuration

"simulator": {
    "conpath": "conf_cpu.json",
    "set": "qiskit"
}

• conpath: Configuration file name for simulation settings
• set: Parameter set within the specified configuration file

There are also configs in ./config folder including: conf_cpu.json for CPU simulations, conf_gpu.json for GPU simulations, conf_noise.json for noise simulations (only support CPU simulation).

Reference

Yuan, H., Long, C. K., Lepage, H. V. & Barnes, C. H. Quantifying the advantages of applying quantum approximate algorithms to portfolio optimisation. Quantum Science and Technology (2026) doi:http://iopscience.iop.org/article/10.1088/2058-9565/ae4a48.

@Article{10.1088/2058-9565/ae4a48,
  author   = {Yuan, Haomu and Long, Christopher K. and Lepage, Hugo V. and Barnes, Crispin HW},
  journal  = {Quantum Science and Technology},
  title    = {Quantifying the advantages of applying quantum approximate algorithms to portfolio optimisation},
  year     = {2026},
  abstract = {We present a quantum algorithm for portfolio optimisation. Specifically, We present an end-to-end quantum approximate optimisation algorithm (QAOA) to solve the discrete global minimum variance portfolio (DGMVP) model. This model finds a portfolio of risky assets with the lowest possible risk contingent on the number of traded assets being discrete. We provide a complete pipeline for this model and analyse its viability for noisy intermediate-scale quantum computers. We design initial states, a cost operator, and ansätze within a binary encoding. Further, we perform numerical simulations to analyse several optimisation routines, including layerwise optimisation, utilising COBYLA and dual annealing. Finally, we consider the impacts of thermal relaxation and stochastic measurement noise. We find dual annealing with a layerwise optimisation routine provides the most robust performance. We observe that realistic thermal relaxation noise levels preclude quantum advantage. However, stochastic measurement noise will dominate when hardware sufficiently improves. Within this regime, we numerically demonstrate a favourable scaling in the number of shots required to obtain the global minimum---an indication of quantum advantage in portfolio optimisation.},
  doi      = {http://iopscience.iop.org/article/10.1088/2058-9565/ae4a48},
}