QC Ware has a different strategy than most quantum software startups. Rather than create a generalized programming platform and offering it to all end users that allows the users to develop their own programs, QC Ware is focused on algorithm development with a specific focus on creating performance speed-ups in near-term hardware. Their team includes several algorithm experts who focus on creating noise resistant algorithms that utilize fewer resources and gate depth so they can utilize NISQ machines as soon as possible. Since one of the biggest deficiencies in today’s machines is the limited gate fidelities, algorithms that have lower gate depths are more able to accurately calculate a good solution. They are developing such algorithms in five problem classes including chemistry simulation, optimization, machine learning, differential equations, and Monte Carlo methods and working with customers who have use cases in these areas.
So even though we often point out the advances in quantum hardware development that is creating more powerful machines to get us closer to achieving quantum advantage, we should also point out that QC Ware and other software companies are also innovating on the software side to enable programs to run successfully on less powerful quantum computers and achieve quantum advantage. There is still a gap today between what we have and what is needed, but think of it as the two sides working together to meet in the middle. When this happens, we will start to see a few organizations announce they are running quantum programs in a production mode. Besides making advances in the algorithms there are other software improvements being made by QC Ware and others to create more efficient compilers and optimized qubit control software to not only lower the gate level resources needed to run a program but also to improve via software the effective fidelity of those gates so they make fewer errors.
The application that QC Ware is researching with Goldman Sachs involves a process called Monte Carlo simulation. These are algorithms which rely on repeated sampling to develop a model that provides insights into risks and pricing of a collection of assets that have a large number characteristics and are interrelated in a very complex way. Classical Monte Carlo simulations are used very heavily in the finance community and finding a workable quantum algorithm that can perform a Monte Carlo simulation could be very valuable because it could either provide the answer in a much shorter period of time or achieve higher accuracy or create much large models that contains more assets and parameters. While Quantum Monte Carlo algorithms have been studied for over more than a decade and could theoretically provide over a 1000X improvement in performance, they would require very large quantum computers that QC Ware does not expect to be available for another 10-20 years. What QC Ware has done is develop a new Quantum Monte Carlo algorithm which they call Shallow Monte Carlo which may only provide a 10-100X performance increase, but could run on quantum computers that they expect to be available in the next 5-10 years. So they have traded off the raw performance for a better timeline. Additional information on QC Ware’s work with Goldman Sachs can be found in a press release available on the QC Ware website here and also a technical paper that describes the new algorithm here.
The collaboration with the U.S. Air Force Research Laboratory (AFRL) has a similar intent of creating less resource intensive algorithms, but in the completely different area of Quantum Machine Learning. AFRL supports both the Air Force and the new U.S. Space Force to revolutionize operations in both the air and space. They are actively researching uses of quantum technology in computation, communications, sensing, and position, navigation, and timing (PNT). They have quantum activities in various centers around the world and support quantum research in a variety of ways including researchers on staff, providing grants, supporting the newly created Innovare Advancement Center, and collaborating with commercial companies such as this one with QC Ware. (You can read more about programs where the Air Force awarded millions of dollars in grants to academic institutions and small businesses in our previous articles here and here.)
The specific program that QC Ware is working on with AFRL is to use quantum clustering and classification algorithms to understand the purpose or mission objective of an unmanned aircraft by observing its flight path. The Air Force can collect data on these aircraft through radar, satellite images, and other means, but to make sense of all the data is a big task. Similar to the previous case with Monte Carlo, there is a classical algorithm which can do cluster and classification called k-means but the classical computers can quickly become overwhelmed with all the data. In 2018, Dr. Iordanis Kerenidis and others, published a paper describing a quantum way of doing this called q-means. The collaboration between QC Ware and AFRL is a two year program that started last year. For the first year, QC Ware has been working to refine and improve the algorithm and reduce the near term resource requirements to it can run on near term quantum computers. It is now substantial more efficient than the one described in 2018. For the next year, the organizations will work together to demonstrate the usage of this algorithms on one or more available machines. For AFRL, the opportunities to utilize quantum clustering and classification algorithms are numerous and they view this project as a test case. Once it is successful, they can look to utilizing the q-means algorithm in many different other places. For more information about this collaboration, you can view the news release on the QC Ware website here.
April 30, 2021