D-Wave Systems, a pioneer in quantum annealing-based computing, today announced significant upgrades to its hybrid constrained quadratic model (CQM) solver that should make it easier to use and capable of solving much larger issues, the company said. The model can now handle optimization problems with up to 1 million variables (including continuous variables) and 100,000 constraints. Additionally, D-Wave has introduced a “new [pre-solver] set of fast classical algorithms that reduce problem size and allow larger models to be submitted to the hybrid solver.
As discussions about using hybrid quantum-classical solutions have intensified recently within the gate-based quantum computer developer community, D-Wave has been actively exploring hybrid approaches for use with its annealing computers. quantum for some time. He introduced a hybrid solver service (HSS) as part of its Jump web access portal and Ocean SDK development kit that D-Wave in 2020. The general hybrid idea is to use conventional compute resources where it makes sense – for example, GPUs perform matrix multiplication faster – and use resources quantum where they add benefits.
HHS also builds on familiar tools and helps meet the nagging challenge of pressing big practical problems onto, relatively speaking, D-Wave’s small quantum systems. Its systems are massive (Advantage has 2,000 qubits, Advantage2 is expected to have 5,000 qubits) compared to current sizes of gate-based quantum computers (IBM is expected to launch a 400+ qubit processor soon). But quantum annealing is a different beast and works differently. In many ways, the comparison is not apples to apples at all.
“No quantum computer will ever be big enough for people to fit their entire application into the computer itself,” said Murray Thom, vice president of product management at D-Wave during a briefing with HPCwire.
Here is D-Wave’s description of the benefits of using its HHS:
- “HSS’s hybrid solvers can accept much larger inputs than those solved directly by the QPU. They are designed to take advantage of the QPU’s unique ability to quickly find good solutions, thus extending this property to larger and more varied input types than would otherwise be possible.
- “HSS solvers are designed to support low-level operational details for the user: troubleshooting with this service requires no knowledge of how to select D-Wave QPU parameters.
- “Different types of solvers tend to perform better on different types of inputs. Portfolio solvers can run multiple solvers in parallel using a cloud-based platform and return the best solution from the pool This approach saves the user from having to know in advance which solver might perform best on a given input, and minimizes the computation time required to obtain the best results.
The new, expanded solver capability, D-Wave reports, “allows quantum developers to better represent business problems, allowing them to more easily and accurately model problems when not all constraints can be satisfied via classic computer logic For example, in an employee scheduling scenario where employees are required to work 8 hours per shift with optional overtime, the solver can now allow a soft “weighted” constraint of
The familiar caveat, again, is that quantum annealing-based computing is better suited to a somewhat narrower set of problems – optimization is high on the quantum annealing suitability list – than gate-based quantum computers. Fault-tolerant, gate-based quantum computers are expected to be able to handle all sorts of computational workloads; However, increasing the size of the fault-tolerant system (number of qubits) and implementing effective error mitigation/correction remains challenging. D-Wave’s quantum annealing system works differently and does not use error correction.
It should be noted that about a year ago D-Wave announced an effort to expand beyond quantum annealing and to develop a gate-based system (see HPCwire cover) and this year the company completed the process of going public through a SPAC. The company has said little about the progress of its portal-based development initiative, but is expected to provide a report at its annual Qubits conference, to be held in January.
Thom highlighted the value of now being able to use weighted constraints with the CQM solver.
“Weighted constraints allow [developers] say, I’m willing to violate constraints in that order of precedence. If they are considering, for example, the overtime constraint for a fleet of delivery vehicles and the vehicle weight constraint, they might say that the weight constraint is going to be difficult. It’s a big priority, it carries a lot of weight. But labor is willing to work two extra hours overtime as it is paid so I can have a softer barrier (weight) to go from an eight hour workday to a 10 hour workday , then a hard limit to go above a 10-hour workday. It gives developers the power to make the solver aware of those nuances in the trade-offs in the customer’s problem,” he said.
The new pre-solving techniques are “integrated” into the Ocean SDK, Thom said: “Pre-solving techniques are used to take an issue instance that a developer has submitted and start analyzing it to say, ‘ There are a bunch of variables here that I can tune in advance. They have no variability. If I modify their value, they are immediately invalid. So I’m going to set the value, then fix them, then pass the remaining part of the problem to the solver. [Pre-solve] is executed each time a problem is submitted; it’s very, very fast, and it basically increases your success.
Of note here is D-Wave’s ambitious commercial engagement program, one of the largest in the quantum community. The fact is that very few quantum applications are running in a production environment. Proof of concept demonstrations and prototype applications abound. The D-Wave quantum annealing approach, in several respects, is simpler than gate-based approaches, which may be one of the reasons why it is apparently more advanced. The company says that at least one organization – Pier 300 of the Port of Los Angeles – uses a D-Wave system for logistics planning in a production environment.
When asked how far real-world apps have come, Thom said, “Well, we’re talking about a whole ecosystem of app development and a range of apps, but the most advanced are such as Port of Los Angeles operations on Pier 300. They are currently optimizing the way cargo is handled on Port 300, which moves materials and moves cargo containers.They have been able to improve handling efficiency freight thanks to the gantry on rubber tires (GTR) cranes by 60% and the turnaround time of trucks picking up these cargo containers by 12%. And they’re making calls to our quantum computer live right now, 24/7.”
It’s a fascinating case story. RTGs are giant and, as you’d expect, expensive to operate. Minimizing their movement was a big goal. D-Wave released a short video on the project, which began in 2018 following the acquisition of Quai 300 for approximately $850 million. The RTG unloads incoming ships, stacked with 20ft containers. Detailed simulations and collaboration with D-Wave and SavantX eventually produced a new system to choreograph the movement of giant cranes and truck traffic loading containers.
D-Wave is an interesting company to watch. It is a pioneer, founded in 1999, which has often been criticized for not being quantum enough and for being limited in the scope of the problems it can tackle. Quantum annealing is indeed different but it has also proven to be powerful. In recent years, other companies have emulated versions of annealing-based computing, whether on quantum devices or classical systems. Now, D-Wave is pushing on both the quantum annealing front and the gate-based and fault-tolerant front, where it says its experience in manufacturing, systems control, software tools and software development apps gives it a head start.