uvic-qplex 0.1.2

an open-source Python library that enables developers to implement combinatorial optimization models and execute them seamlessly on multiple classical and quantum devices using different quantum algorithms.

⚠️ This library is currently under development.

QPLEX is an open-source Python library that enables developers to implement combinatorial optimization models and execute them seamlessly on multiple classical and quantum devices (using different quantum algorithms). Our solution automatically handles the adaptation of the optimization model to the specific instructions of the target quantum device's SDK. The library supports unconstrained and constrained problems, as well as binary, discrete, and continuous variables. QPLEX uses DOcplex as a modeling API allowing the creation of optimization models using the exact same syntax as this base library.

The motivation and technical details about this project are presented in the following paper: https://arxiv.org/abs/2307.14308

Getting Started

QPLEX is not a self-managed solution, hence each user has to bring their own quantum provider access token in order to use quantum devices. All tokens must be associated with environment variables within the system.

For using devices from IBM Quantum the following env variable has to be created: IBMQ_API_TOKEN

For using devices from D-Wave the following env variable has to be created: D-WAVE_API_TOKEN

In Amazon Braket's case the user must have the AWS CLI configured properly and should have access to the AWS Braket service in the cloud. Follow this link for more information.

The QModel module provides all the necessary functionality for creating and executing a combinatorial optimization problem. The following example illustrates how to build and run the knapsack problem with QPLEX. More examples are provided within the "examples" folder.

from qplex import QModel
from qplex.model import Options
# Problem definition
weights = [4, 2, 5, 4, 5, 1, 3, 5]
values = [10, 5, 18, 12, 15, 1, 2, 8]
max_weight = 15
n = len(values)

# Build the model
knapsack_model = QModel('knapsack')
x = knapsack_model.binary_var_list(n, name="x")
knapsack_model.add_constraint(
    sum(weights[i] * x[i] for i in range(n)
        ) <= max_weight)
obj_fn = sum(values[i] * x[i] for i in range(n))
knapsack_model.set_objective('max', obj_fn)

It is worth noting that the syntax used by QPLEX is exactly the same as the one used by DOcplex. This facilitates the integration of our solution into classical workflows.

QPLEX provides a new functionality within the solve method; it allows the user to specify how the problem will be executed, either through classical or quantum resources. If nothing is passed to the method, classical resources will be used.

When calling the solve method with "quantum" resources, the software will automatically select the most appropriate algorithm and quantum hardware backend for the defined optimization model. However, the users have the option to specify, as kwargs, which algorithm, provider, and backend they want to use for the execution.

The following is the complete list of kwargs supported for the solve method:

Argument	Description	Default value
provider	The quantum hardware provider	"d-wave"
backend	The specific quantum device	Calculated
algorithm	The quantum algorithm to use	"qaoa"
ansatz	The ansatz circuit for VQE	Layered Ansatz
p	The p value for the QAOA algorithm	2
layers	The number of layers for the VQE algorithm	2
optimizer	The classical optimizer	"cobyla"
tolerance	The tolerance value for the optimizer	1e − 10
max_iter	The maximum number of optimizer iterations	1000
penalty	The penalty constant used for the QUBO	Calculated
shots	The total number of shots	1024
seed	The execution random seed	1

Supported Algorithms

QPLEX currently supports the following algorithms:

• Quantum Annealing (QA)
• The Variational Quantum Eigensolver (VQE)
• The Quantum Approximate Optimization Algorithm (QAOA)

Each of these has its own set of hyperparameters that can be passed as kwargs to the solve method. Refer to the previous table for the complete list.

execution_params = {
        "provider": "braket",
        "backend": "simulator",
        "algorithm": "qaoa",
        "p": 4,
        "max_iter": 500,
        "shots": 10000
    }
knapsack_model.solve(solver='quantum', Options(**execution_params))

Algorithms to be supported in the future:

• The Quantum Alternating Operator Ansatz (QAOAnsatz)
• Warm-start QAOA (WSQAOA)
• CVaR QAOA (CVar)

Supported Quantum Providers

The quantum providers supported by the library are:

• D-Wave
• IBMQ
• AWS Braket

Each of these has its own set of backends (i.e., quantum computers) and can be selected by providing a value to the "backend" key-word argument in the solve method. If no argument is specified, the library will automatically select the backend with the shortest queue and enough qubits to handle the formulation. Additionally, it is possible to use the argument simulator as a backend to use each provider's local simulator.

knapsack_model.solve(solver='quantum', provider='ibmq', backend='ibm_perth')

knapsack_model.solve(solver='quantum', provider='braket', backend='device/qpu/ionq/Harmony')

Contributing

If you are interested in contributing, please check the issues page and select the one you want to address. Afterward, fork the repository and create a new branch with the issue's number. Make sure to push all you changes in a single commit with a descriptive message. If the issue description is not clear, feel free to create a comment requesting more information.

Feature requests, bug reports and feedback comments are highly appreciated! 😃