QxBranch at AQC 2015 (ETH Zurich)
On Monday, June 29, QxBranch Director of Software Engineering Mark Hodson was invited to present at the international Adiabatic Quantum Computing (AQC) conference being held at ETH in Zurich, Switzerland. The conference brings together luminary researchers from different communities to explore the landscape of AQC architectures and algorithms, and methods for realizing them.
Before Hodson formally presents his findings entitled, "A Novel Embedding Technique for Optimization Problems of Fully-Connected Integer Variables," he gave an overview of his research and its potential significance.
Q: Describe your research in a 'nutshell.'
Our research focuses on two practical aspects of solving problems on the D-Wave Two Adiabatic Quantum Computer when the variables are integers (multi-valued), rather than binary (two-valued).
Q: Why does it matter?
We show that it's possible to solve practical multi-valued problems on the D-Wave Two, and we identify areas for future collaborative improvements. It involves a novel application of existing embedding methods to multi-valued problems and a new efficient and effective method for calculating penalty coefficients for any problem type (including binary).
Q: What is the outcome you envision?
QxBranch is now positioned to inform QC architecture development by identifying opportunities for improved outcomes for client-focused problems. We have identified a pattern of hardware use that could be optimized by hardware vendors for improved client outcomes.
Down the line, the next step could be to consider how multi-level problems might be sub-divided to improve dynamic range as part of a divide-and-conquer engineering strategy, but we haven't committed to fully exploring it at this point.
Q: How does this work translate to QxBranch's clients?
Any client with a multi-valued optimization and constraint problem, such as finance portfolio positioning problems that operate in quarter-percentile increments, should be excited about the potential for AQC architectures to provide benefit to their business. The general AQC community should also look forward to the broad applicability of the novel Hamiltonian penalty calculation technique that QxBranch has developed.
Q: Any limitations?
Right now, our technique has limitations on the current generation hardware. To be practical it requires mores operational, coupled qubits as higher operational range of biases and couplings that can be programmed onto the D-Wave Two.
However, we're positioning ourselves for the future. Planned improvements to AQC control architectures are expected to yield improved dynamic range, and new fabrication techniques are expected to improve qubit yields.
New commercially available AQC systems are already in development. By researching and assessing performance problems with practical client applications now, QxBranch can analyze and inform technology development pathways that provide the greatest and earliest possible benefit to our clients.
Mark Hodson is the Deputy Chief Technology Officer for QxBranch and has established a framework of software libraries and reusable components which provide engineers and scientists on the QxBranch team with a systematic way to solve real-world problems on the D-Wave Two. He has also lead many aspects of the projects undertaken by QxBranch. Over the last 17 years, he has made substantial contributions to knowledge in the fields of modeling and simulation, systems engineering and medical image processing.