Various Types of Solvers and Problems They Can Solve. Credit: D-Wave

One of the challenges for providers of quantum computing technology is to make it easier for end users to take advantage of its capability by providing tools that can help abstract the problem and let the subject matter experts specify the problem at a higher level. D-Wave has been providing a tool for some time that they call a Solver. The initial version would accept problems with binary input, only 0 or 1. Later they introduced a version that could accept problem inputs with discrete variables, e.g. 0, 1, 2, 3, etc. In October of last year, they introduced a solver called the Constrained Quadratic Model (CQM) that allowed a user to add constraints to their problem and last May they upgraded the CQM to also handle continuous as well as discrete variables. All of these upgrades made it easier for an end user because more of the steps of reformulating the problem into a form that could be run on the physical machine can be done by software instead of manually by the end user.

Now, D-Wave has upgraded the CQM solver again by offering the ability to specify weighted constraints. Think of it as having “soft” and “hard” constraints. An example provided by D-Wave was for a scheduling application where the soft constraint for a worker was a shift of 8 hours or less. The worker can work more, but overtime pay will be required up to a hard constraint of a 12 hour shift. The CQM solver works in a hybrid mode and another feature being introduced is what they call presolve techniques which are implemented on a classical computer. The presolve techniques will remove unnecessary variables and constraints to narrow down the problem size and streamline problem formation. Again, this process can be performed automatically and it enables larger models to be submitted to the hybrid solver.

For additional information about these new features in D-Wave’s solver, you can see a press announcement here, a benchmarking report here, and sign up here to attend a webinar on November 17, 2022 that demos the software.

November 3, 2022