QUBO-for-Qubit-Allocation

Uses the QUBO formalism to generate a cost function whose solutions are allocations of logical qubits from a quantum circuit to the physical qubits of a hardware graph. Solutions from the cost function are obtained using simulated annealing, as implemented in D-wave's neal. An investigation of how effective this method is at generating initial allocations is available here (https://arxiv.org/abs/2009.00140). Code was written by Bryan Dury (@bdury), but the paper is a product of both @bdury and @glassnotes.

Installation

The specifications of an Anaconda environment are available in the environment.yml file. To reproduce the environment install Anaconda and run the following command:

  conda env create --file environment.yml

Note that the project specifies Qiskit v.0.20.0, but uses an unstable build of qiskit-terra for generating the data used in the paper. If you don't want to try and replicate our results, just follow the above instructions. Otherwise, if the current available version of qiskit-terra is > v.0.15.1, use the most recent version available. If this is not the case please follow the guide here to install qiskit-terra from source.

Also, there are some functions under benchmark.py that use pytket which is only available on macOS or Linux distributions, and I developed this mainly on Windows, so it is not included in the environment file. If you want to use those functions please install pytket using pip.

Working with the Code

There is a Jupyter Notebook with a minimal working example that leads you through how to generate QUBO initial allocations under the folder examples. The source code itself contains many other useful functions to work with the data that is produced from each simulated annealing run for each set of circuits - please have a look if you are interested in working with the data.