Report: Quantum computers are now better equipped to solve optimization issues

It’s widely known that quantum computers are well suited for solving optimization problems, an application which is a front-runner for showing performance improvements over classical computation in the near future. Until now, however, it was unknown how much universal quantum computers had improved at this task given the rapid development of new and more powerful hardware. Is quantum optimization on a path towards supremacy?

A new report by Agnostiq indicates that performance has indeed improved, as demonstrated by its financial portfolio optimization benchmarks on a selection of gate-model quantum computers. Agnostiq’s method is to benchmark the quality of portfolios obtained after the completion of many optimization tasks on a selection of superconducting and trapped ion quantum computers. Specifically, it constructed four representative problem instances using historical stock data in the tech market and solved them using (1) five IBM superconducting chips, (2) one Rigetti superconducting chip and (3) one IonQ trapped ion chip.

A crucial parameter for increasing the performance of the algorithm is the circuit depth. Circuit depth is the number of operations (think AND / OR logic gates in classical computing) used to solve the problem. Simply put, as circuit depth increases, the quality of the quantum solution to the optimization optimization problem should increase.

However, in the current era of noisy hardware, increasing the circuit depth too much leads to decreasing performance as the effect of noise takes hold. There is therefore a “sweet spot” located at the point before performance decreases, which, as Agnostiq has found, has improved for the most comparable of problems.

Agnostiq also shows the first deployment of a more advanced version of the algorithm to hardware which also shows improving solution quality with circuit depth. An interesting observation is the non-alignment of its application-specific performance metrics with more general metrics like quantum volume. That is, higher quantum volume did not always translate to higher application performance, showing that more research is needed to suitably quantify the performance of quantum computers.

Read the full report by Agnostiq.