Quantum Optimizations at Qubits, Day 2
D-Wave’s quantum annealed Success Story
The following notes represent an unofficial meeting minutes of sorts, recording the discussions presented at the Qubits Conference hosted by D-Wave Inc June 18, 2024 in the beautiful city of Boston, MA. (An writeup for Day 1 of the conference was shared in a separate publication linked again below.) The conference was associated with quantum annealing, the adiabatic form of quantum computation that D-Wave has been pioneering for optimization applications going on 25 years, and scope of the proceedings were more aligned with the conventions of a commercial user meetup (unlike some of the research venues that this blog has previously visited). The technical breadth of this writeup will fall on the lighter side — there are more traditional academic venues being held around this domain like the Adiabatic Quantum Computing conference (AQC) if you are research inclined. As will likely become apparent from the reading, it is an exciting time for the commercialization of quantum computing and for D-Wave in particular given their extremely differentiated contributions to the field of high performance computing.
The format of this writeup will quite simply adapt various presentations into a “key takeaways” form of prose, following the published agenda and sequence of speakers from the conference — which the author has attended in person for a second year running. The author has also sat several other forms of training offered by D-Wave and so considers himself literate in their product line, which includes the Ocean API for high bandwidth / low latency optimization, the Leap cloud service for quantum solver access, and the Launch professional services available to support organizations along their journey from developing proofs of concept all of the way to scaled deployment.
As a disclosure, please note that the author has active investing interests in D-Wave through their publicly traded $QBTS equity. He is a long term / passive investor and has no intent of using this publication for purposes of “trading”. He offers this writeup as an attempted contribution to the field as well as a thank you to the employees for their kindness and hospitality in these events.
Great so without further ado:
Opening Keynote: Quantum Trends Powering Organizational Decisions
Brian Lenahan, Chair, Quantum Strategy Institute
Coming from a banking background and with recent authored books like Quantum Boost, Quantum Excellence, and On the Shoulders of Giants, Brian wishes to support business professionals as they evaluate the potential of quantum technologies for their field. He founded the Toronto based Quantum Strategy Institute as an international philanthropic think tank of sorts, staffed by volunteers with respectable professional backgrounds looking to support industries, organizations, and institutions as they develop road maps for their integrations of quantum compute.
He finds that many organizations are actively making significant investments into emerging technologies but have somehow fallen into the misconception that AI should be considered the highest priority and quantum technologies are more appropriate for exploratory initiatives. He says the business reality is that establishing some form of competitive advantage from AI technologies would likely require large upfront training costs to initiate, while quantum technologies can be scaled up to production much quicker.
The Quantum Strategy Institute (QSI) is part of a consortium of like-minded industry and regional associations like QED-C, QuiC, QIC, and QStar. Their output includes reports on topics like “how can countries prepare themselves for quantum vulnerabilities in cryptography?” This turns out to be an urgent issue for the field, and is one reason that the CTO of a prominent international coalition of nations is recommending that the timeframe for rolling out quantum to industry should be 2–5 years maximum.
Evaluating the performance of competing hardware conventions has traditionally been led by top vendors seeking to publish and promote their own forms of benchmark metrics that play to whatever are the particular strengths of their particular conventions of hardware. The preference of QSI is to step back from such microscopic benchmarking conventions and consider performance from a broader ecosystem sense, particularly as many clients have varying levels of domain comfort which make their interpretation of such esoteric forms of comparison inconsistently grounded. Sometimes new customers may see cybersecurity marketing of the like “it is better to fight quantum with quantum” where even the use of the word quantum is distracting and counterproductive to customers without domain experience. It may be better to save such language for when they establish a minimum level of comfort, as each organization will experience their own roadmap of learning. Common language and terminologies for the industry have yet to be fully established, as have expectations of an appropriate return on investment.
Application of Quantum Annealing to SAS Advanced Analytics
Bill Wisotsky, Principle Technical Architect, SAS
Bill has followed a fascinating career trajectory along his journey towards quantum computing. After withdrawing some time ago from police training due to a visual acuity exam he enrolled into a neuroscience graduate school, where he began learning about how photons are absorbed by the retina. As he began digging into this subtopic he found in the literature discussions of those basic quantum experiments that we all know so well, the double slits and a boxed in cat, which is a channel of interest he followed adjacent to his more recent career with SAS, a respected international consultancy for data and software analytics. He now finds himself on a team of professionals seeking to develop a roadmap for SAS’ integration of quantum computing technologies into the firm.
SAS is in a good position to evaluate D-Wave’s quantum annealing technologies as they have a long established team of in-house optimization experts. They found that some of D-Wave’s combinatorial optimization framings were similar to their experience, which made the learning curve a breeze, sufficient for them to establish that annealing was the only real technology available today capable of running real world scale optimization problems for quick turnaround of customer applications.
As a demonstration he shared an evaluation of the “kidney exchange problem”, an application whose developer had been awarded a Nobel prize in 2012. The framing was designed for a healthcare setting where one wishes to establish chains of kidney exchange patients in context of variable compatibility between patients and a limited pool of donors. Evaluation required assembling a graph of compatibility between patient donor pairs with scoring evaluated for unique multiway swap chain configurations. When such a setup was channeled to a classical optimization solver at limited scales, feasible configurations with low compatibility scores became accessible within seconds, but in order to approach the optimal solution it required around an order of magnitude longer solver evaluation time. The D-Wave solver was able to access a sampled distribution of the solution space in near constant time.
Quantum for Climate and Sustainability
Annarita Giana, Senior Scientist, GE Vernova
For those who haven’t kept up with the news, the historic industrial conglomerate General Electric, whose ancestry dates back to the days of Thomas Edison, had a form of reorganization earlier this year where they channeled their businesses aligned to healthcare, aerospace, and energy segments into separate organizations. Annarita was here representing GE Vernova, which inherited power generation and energy industry aspects of that portfolio. She works on optimization and control with a team of about 300 researchers in upstate New York who are looking to mitigate the impact of carbon intensity in our climate alongside other product line initiatives like improving wind turbine asset performance.
She highlighted that the National Science Foundation recently hosted a workshop for climate and sustainability initiatives which issued an extensive report highlighting optimization technologies that could be applied to domains like chemistry, asset modeling, and demand forecasting. She gave examples of several other sustainability objectives that could be good candidates to serve as future targets for integration of performant optimization technologies, like the allocation of renewable energy resources, carbon capture deployment planning, supply chain processes, urban planning and infrastructure design, water resource management, biodiversity conservation planning, circular economy practices (like minimizing waste from recycling), natural disaster response planning, resilient agriculture planning to adapt to changing climate, and operating a smarter electric grid.
Quantum Annealing and Power Consumption — A Case Study in Mobile Telecommunications
Paul Warburton, Professor, University College
In a field with headlines dominated by feats of nanoscale engineering, Paul’s research zooms way way out, climbing the layers up a ladder of abstractions to a much grander scale in order to identify those macro segments of our economy where the deployments of quantum annealing technologies may be of the most immediate benefit to our society. He thinks he may have found just such an application in the field of mobile communications, but fleshing it out requires putting on an engineer’s hat and assessing operating envelope requirements and capabilities all the way from first principles.
Why would mobile communications merit the prioritized integration of annealing technologies? Consider the global energy demand of data centers we have already noted earlier in the conference. Mobile communications is also large enough to move the needle of the global trends and has the bonus of a durable operational profile. Paul demonstrated a full derivation in which he inspected what may be considered the current bottleneck of 5G wireless implementation, and by considering the number of mathematical operations per second and corresponding energy draw of existing solutions he found that an alternative annealing based operation, whose energy demand is primarily associated with the fixed profile of a dilution refrigerator on the order of 25kW, would have potential for a material cost savings from electricity demand in this form of scaled deployment — with only a few remaining technical hurdles to climb.
He further suggested that any predictable computational load profile with a power demand in excess of a dilution refrigerator should be evaluated as a potential candidate for realizing those energy savings available from a quantum computer.
Exploring the Optimization of Key Business Processes Using Quantum Computing: Survey
Bob Sorensen, SVP Research, Hyperion Research
Bob has been following the field of super computing for nearly 40 years now. His offices conducts surveys for research reports intended to inform industrial intelligence initiatives that may support potential users or investors evaluating economics of the field. His most recent report, which was completed shortly before the conference, was conducted without the input, influence, or financial support of D-Wave and is available for purchase from Hyperion Research.
It would not be appropriate for this writeup to share specific key figures from the report, as Hyperion is a commercial vendor whose key sources of revenue are associated with selling access to such information, but this blogger expects that it would sufficiently convey the significance of their findings to indicate that of the hundreds of firms that responded to their survey, many were expecting to spend millions of dollars on emerging technologies for performant optimization within the next 18 months, and on average were projecting ROI on the order of ten times their investment.
Transforming Airline Operations with Quantum Annealing
Jefferson Smulin, Director of Emerging Technology, Unisys
Unysis has followed the progression in airline and airport operations for nearly four decades, so they are qualified to be considered experts in the field. Jefferson began partnering with D-Wave about a year ago to build solutions for a high value application in airline operations associated with cargo allocation between freight and passenger airplanes.
With our modern global marketplace for shipping and receiving goods, and particularly with changes to our economy that arose from the pandemic years, airlines are faced with shifting expectations and a growing demand for just-in-time delivery and logistics solutions. Consider that as packages arrive to an airline and are routed towards different modes of transportation, in some cases there may be unexpected challenges from edge cases like weight distribution on a smaller plane, urgent deliveries, and other occurrences of the unexpected. Solutions that rely on classical computing based solvers can do a lot, but in order to adapt in real time to unexpected operational challenges, quantum annealing is uniquely capable of real time adjustments to operational plans.
In a field like air traffic where those rare edge cases are the ones that end up causing the most heartache and making the evening news, those optimization solutions being developed by Unisys that can leverage real time solvers available from D-Wave will be uniquely capable of ensuring the highest consistency of passenger satisfaction, and that is the end goal isn’t it?
Harnessing D-Wave’s Binary Quadratic Model Solver to Deliver Results for Drug Discovery
- Bill Shipman, Founder and CTO, POLARISqb
- Maurice Benson, Principle Software Engineer, POLARISqb
Current state of the art in drug discovery and design typically involves researchers conducting multi-year evaluations and experiments costing millions of dollars to implement. Evaluating a limited set of candidate molecules, they may find themselves iteratively solving a sequence of desirable chemical properties in order to refine designs of protein binder molecules over repeated experiments and simulations.
Bill is looking to make such arduous research processes obsolete. Based out of Durham, NC, his firm POLARISqb is leveraging D-Wave solvers in a working implementation that allows conducting candidate protein binder molecule searches on a vast scale, evaluating billions of candidate drug molecules all at once in a short duration optimization in comparison to prior methods that used instead classical solvers to inspect thousands over prolonged solver times.
By considering candidate molecules as an assembly of tractable molecule fragments, they can in parallel constrain solutions to only include those molecules that are known to be synthesizable and adaptable to the specific biological process in which they wish to fit a protein binding. They consider it a strong validation that in addition to novel drug candidates, their solutions often return previously discovered molecules that are known to be performant to some biological mechanism.
The D-Wave annealer allows solving an incredibly large chemical space infeasible to existing solutions. Maurice noted in closing that “Quantum utility is available today, don’t get lost in definitions, go with the better outcomes.”
Quantum Annealing for Therapeutic Peptide Design
Andrejs Tucs, Researcher, University of Tokyo
It could be considered a society scale problem that many parts of the world continue to conduct inappropriate use of antibiotics at scale. This is known to contribute to our limited set of medications becoming less efficient, and enabling bacteria strains with antimicrobial resistance. The United Nations last year issued a report serving as a warning that at current trends, it is projected that in time even mild infections may become deadly again. In the millions of years they have had to evolve, bacteria until now had yet to develop resistance against antimicrobial peptides. Make no mistake, this is a global threat with global consequences.
Andrejs proposes that we begin looking for new and more specialized antimicrobial peptides. Given that amino acids form chains in which their unique 3D structures arise to unique variations in features and functionality, it is reasonable to expect that there should be a huge space of yet to be discovered peptides that could serve such use. As peptides have difficult to simulate characteristics in which even small changes in composition may critically affect their functionality, he wishes to apply the power of D-Wave’s quantum annealers to explore multiple property landscapes simultaneously.
Prior work has sought to leverage variational auto-encoders, a form of generative AI, that were trained by compressing molecule representations to a binary latent space and then attempting to reconstruct their full representation. His work has sought to extend these conventions with a quantum unconstrained binary optimization framing (QUBO) to relate this model to an adjacent representation of material properties. Such setup allows them to account for multiple objectives of material properties as simplified to a single relationship that falls along some Pareto frontier of tradeoffs between objectives, serving as a form of simplification to the solver formulation. MOQA, his pipeline for biological sequence design with multi objective optimization by quantum annealing, has identified peptide designs with better targeted antimicrobial properties than existing baselines.
Application Demo
Catherine Potts, D-Wave Senior Technical Advisor, D-Wave
Serving as a technical advisor to the D-Wave user community, Catherine introduced herself with some folksy tales of growing up in Montana, where the Rocky mountains views are pristine, the Yellowstone river rolling along the countryside full of hungry fish waiting to be caught by the casts of a fly fisherman, and the highlights of the week are often a home grown ice cream shop full of tasty goodness on a hot summer afternoon.
She presented a demonstration allowing the D-Wave optimization solvers to run real time updates to the local employer workforce scheduling, as is especially needed when everyone wishes to take those perfect summer afternoons off for a walk along the countryside — even though some of us have to stay behind and work! This and other demonstrations of solver framings are available from the D-Wave technical advisor team to support whatever adaptions you may wish for your hometown needs.
Fast Tracking the Quantum Journey of Industries: LTIMindtree’s Jumpstart Approach
Vijay Rao, Principal Director and Research Lead, LTIMindtree
At the Mumbai, India headquarters of LTIMindtree, a global consulting firm offering digital solutions to diverse industries, Vijay has been incubating several new technologies and solutions designed to support a wide range of use cases that will leverage D-Wave annealers. He is already using in production annealer based forms of portfolio optimization that adapt equity theory or Markowitz based option trading to operate real time trading strategies. Using traditional forms of budget and risk constraints paired with return objectives, they tap the wide scale of annealer bandwidth capacity to evaluate large volumes of investment candidates for return profiles, with diversification across assets and asset classes to dampen many forms of risk.
He sees a expansive range of further opportunities that are candidates for annealing solutions, and listed several more including employee scheduling (particularly for the hospital and nurse setting), oil refinery operations, capacity weighted vehicle allocations, and many more.
Vijay offers LTIMindtree as a quantum ready partner available to fast track quantum value to diverse industries. As every company has to go through their own journey of quantum awareness, competence, and value creation, he suggests “getting to the future faster, together with D-Wave.”
Performance of Annealing Quantum Systems: Theory Versus Reality
Catherine McGeoch, D-Wave Principal Scientist Emeritus, D-Wave
Having served on the D-Wave benchmarking team for over a decade, Catherine has had an enjoyable career evaluating performance of technological innovations. Along the way she has seen all manner of questions asked on how to quantify performance. One of the most important remains “When can annealing outperform classical heuristics?” It has been difficult to offer a precise answer to this question until very recently, when she partnered with a colleague to write a paper about quantum utility.
Some patterns often show up from literature review. It helps with such evaluations to draw lines along the distinctions between physical and logical qubits. Similarly for annealers it helps to distinguish between anneal time and access time and also to consider the differences between some physical graph of qubits versus any modifications associated with aggregated chains of qubits for purposes of application encoding.
When evaluating the latency of a quantum annealing application, you have to consider that although the new fast anneal conventions may support the scale of a few nanoseconds per sample, the surrounding access and readout time may be closer to the several millisecond range. She suggested that if you might like a way to envision, you could assume that a typical encoding / anneal / readout might take a time scale of such multi millisecond range. That being said, once you have a problem framing into the encoding form, there is no reason that you can’t draw as many nanosecond scale samples as you like, which in comparison to the readout time will have negligible impact to the time budget. Consider this in comparison to some classical solvers in which assessing a distribution of samples is inaccessible to a user at any reasonable time scale.
Restating for clarity, the benefit of quantum annealing comes not just from gaining access to a desirable optimization solution in some short amount of time, it likewise arises from having access to inspect the distribution of sampled solutions — potentially in a timeframe comparable to a classical solvers initial sample. After all, classical heuristics like simulated annealing start at a random position in a fitness landscape and use random fluctuations to find “downhill paths”. One of the implications is that for any given heuristic, there will always be some input where a classical solver will get stuck in some local minimum of the fitness landscape, which is why they require hyperparameter tuning, another source of advantage for quantum annealing as it exhibits nearly stable behavior across a wide range of input and parameter configurations.
Neglecting these surrounding matters, if you wish a straight forward answer to the straight forward question of when we can expect quantum annealers to outperform classical heuristics for evaluation of a single approximate optimization solution quality, Catherine’s results clearly identified that quantum annealing has inherent advantages that can be counted on for those largest problem sizes requiring the fastest returned solutions.
Catherine reminds us further that similar to the energy demand, the annealer access time has appeared to remain comparable through various generations of hardware, giving reason for optimism of further advantages against classical as future generations of quantum annealers become available.
Short-Depth QAOA Circuits and Quantum Annealing on Higher-Order Ising Models
Elijah Pelofsky, Research Student, Los Alamos National Laboratory
One should not leave this reading with the impression that quantum annealers are the only form of quantum computer algorithms available to support optimization applications. Gate model quantum circuits likewise have conventions available which may approximate an optimization solution, although they follow a different mechanism to do so.
Elijah sought to benchmark these conventions, comparing the performance of D-Wave’s quantum annealers against an algorithm available to gate based circuits commonly known as the quantum approximate optimization algorithm. To do so he accessed a recent circuit available from IBM released under the name “IBM Washington” that was configured with 127 qubits in a hexagonal pattern.
Part of the challenge of such algorithmic benchmarking conventions is attempting to find some way to conduct a like-for-like comparison in a manner that does not unduly leverage the inherent hardware advantages between different conventions, like those advantages available to the >4800 qubits sampled from the D-Wave Advantage platform system in this benchmark compared to the IBM circuit’s 127 qubits. Elijah found that one way to conduct such hardware agnostic algorithmic benchmarking was to design synthetic optimization problem formulations that took maximum advantage of the qubit counts and coupling configurations available on the more limited IBM hardware, which he found were capable of being redundantly encoded onto the given scale of D-Wave Advantage annealer hardware, such that each sampled solution from the annealer produced six concurrent unique solutions into comparison to a single solution sampled from the IBM hardware.
Another challenge of the like for like comparison involved the forms of tuning needed for the IBM gate model operations. In order to make these more comparable he added additional tuning to the annealer schedule characteristics so that they could have comporable runtimes.
With regards to the results of Elijah’s benchmarks, I don’t think this blog would be an appropriate venue to detail them explicitly, but it should be sufficient to convey that the distribution of sampled optimization solution quality appeared to reinforce the interpretation of a D-Wave quantum annealer advantage.
Financial Applications for Quantum Computing that can be Instituted Today
Ethan Krimins, CEO, Quantum Research Sciences
Coming from the peaceful cornfields surrounding Purdue’s Lafayette, IN campus, Ethan’s surroundings are not what you traditionally think of in the world of high stakes finance. His work with D-Wave’s annealers includes those domains of portfolio optimizations and investing theory that are more commonly associated with Manhattan skyscrapers and long hours wearing uncomfortable collared shirts and ties.
He offered with this presentation sort of an introduction to the types of applications available to leverage quantum annealing in the world of finance. As portfolio allocations between assets like stocks and bonds may be considered an obvious application for such derivations, it should be remembered that these types of applications are not far removed from those decisions that board members may make when balancing a corporation’s capital structure or allocating resources between business initiatives. Choosing a set of stocks need not be tied to a specific number of securities either, consider that in some cases reducing the number of assets to those with unique but known circumstances could actually benefit risk profiles. A reader may consider such discussions as an unlocked door opening to a world of these and similar applications, if you are interested try ringing the bell and Ethan may be available to answer if you’re lucky.
GMV Leading D-Wave to Space
Ana maria Sánchez Montero, Quantum Section Head, GMV
Ana shared with us the story of GMV, a European space contractor now active in twelve countries and counting. Founded in 1984 by a professor passionate about flight mechanics, his small team won an international competition for a space operations center. Forty years later, they have grown to 3,000 employees and are orbiting the upper stratospheres of high tech. As with any space company they have adjacent operations supporting defense and security, in parallel they are expanding into AI, and they strongly believe that quantum computing is the future of their field — in fact they are acting on that belief.
Their highest profile application for the D-Wave annealers appears to be satellite mission planning, which involves an exponential number of real time decisions associated with balancing objectives and constraints. One example she gave leveraged the D-Wave annealers to sample from a search space of an astronomical number of configurations. That’s pretty big!
Emcee Closing Remarks
Murray Thom, D-Wave VP Quantum Technology Evangelism, D-Wave
It wouldn’t do justice to the impressive nature of this conference without acknowledging Murray, the acting master of ceremonies for a second year running, and whose 21 years with D-Wave is an uncommon resume in a field where the majority of workers are just getting started.
He offered to the crowds in closing the reminder that the famed physicist Richard Feynman once famously said:
Understanding something doesn’t make it less wonderful.
It makes it more wonderful.
Day 1 essay shared in a preceding post linked here.
For more essays please check out my Table of Contents, Book Recommendations and Music Recommendations.
© Nicholas Teague 2024, all rights reserved
