Allowing Quantum Developers to Exploit the Power of Quantum Turing Machine Models
Introducing Control Flow in Qubit Allocation for Quantum Turing Machines
Key points…
+ Different platforms for quantum computation are currently being developed with a steadily increasing number of physical qubits. To make today’s devices practical for quantum software engineers, novel programming tools with maximal flexibility have to be developed. One example to extend the applicability of quantum computers to more complex computational problems is quantum control flow. The concept of control flow allows for expanded algorithmic power of the programming language in the form of conditional statements and loops, which alinearly-executed program is incapable of computing. In this work, we introduce a framework to reconcile the non-deterministic properties of quantum control flow when allocating logical qubits from a given quantum circuit to a specific NISQ device in the pre-processing and compiling stage. We consider the respective connectivity and fidelity constraints, with the goal of reducing the expected error rate of the computation. This work will allow for quantum developers and NISQ devices together to more efficiently exploit the compelling algorithmic power that the quantum Turing machine model provides.
Our implementation introduces the first practical techniques to optimally allocate quantum programs onto NISQ devices, and will be seamlessly integrated with QPU execution when devices begin to support intermediate measurement of physical qubits.
+ Future research will test the performance of our implementation for a variety of control flow circuits across several devices/simulators and architectures and explore how the rate of convergence of the SSO in the experiments changes based on properties of the circuit/back-end, as this would allow us to potentially predict how many trials are sufficient to capture the overall distribution of the circuit. We also seek to address the question of how one may consider debugging a quantum program, that includes control flow commands, or how we can further reduce the noise of the computation by e.g. using randomized compilation approaches, or other circuit optimizing routines. Another interesting route for improvement would be by using algorithms introduced for machine learning/data science application, such as a Bayesian optimizer considering the outcome of already performed quantum measurements to predict the weights of individual building blocks on the fly.
+ Enabling quantum control flow on real devices is a significant and essential milestone for the field of quantum computing, as state-of-the-art quantum algorithms, hybrid quantum-classical algorithms, and quantum error-correcting codes, will benefit from control flow and Turing-complete instruction languages.
Content may have been edited for style and clarity.
