Russian scientists use supercomputer to probe the limits of Google’s quantum processor

Representation by an artist on the Google processor. Credits: Forest Stearns, Google AI Quantum Artist in Residence

CPQM’s quantum information processing laboratory worked with the high-performance computing team at CDISE “Zhores” to emulate Google’s quantum processor. By recreating the data noiselessly using the same statistics as Google’s recent experiments, the team was able to highlight the subtle effects of Google’s data. This effect, called lack of accessibility, was discovered by the Skoltech team. Past work .. The numbers confirmed that Google’s data was at the end of the so-called density-dependent avalanche. This means that future experiments will require a lot of quantum resources to perform the optimization of the quantum approximation. The results will be published in the main journals specializing in this field. quantum.

From the earliest days of numerical computations, emulating quantum systems seemed very difficult, but the exact reason is still actively researched. Yet this difficulty apparently inherent in the classical computer that emulates quantum systems has prompted some researchers to turn history around.

Scientists such as Richard Feynman and Yuri Manin were themselves using unknown components as computational resources in the early 1980s, which seemed to make it difficult to emulate quantum computers using classical computers. I guess I can do it. For example, quantum processors must be good at simulating quantum systems because they are run by the same basic principles.

These early ideas eventually led Google and other tech giants to create prototype versions of the long-awaited quantum processor. These modern devices are error prone and can only run the simplest quantum programs, and each calculation must be repeated several times to average the errors to ultimately form an approximation.

Some of the most studied applications of these modern quantum processors are quantum approximation optimization algorithms, or QAOA (pronounced “kyoo-ay-oh-AY”). In a series of spectacular experiments, Google used a processor to study QAOA performance using 23 qubits and three adjustable program steps.

In summary, QAOA is an approach aimed at approximately solving optimization problems in a hybrid configuration composed of a classical computer and a quantum coprocessor. Typical quantum processors, such as Google’s Sycamore, are currently limited to performing noisy and restricted operations. Hybrid configurations alleviate some of these systematic limitations and allow quantum behavior to be restored and utilized, making approaches such as QAOA particularly attractive.

Skoltech scientists have made a series of recent discoveries related to QAOA. For example, see the article. here .. Most notable is the effect which fundamentally limits the applicability of the QAOA. They show that the density of the optimization problems, the ratio of the constraints to the variables, acts as a major obstacle to the realization of an approximate solution. Overcoming this performance limitation requires additional resources for the operations performed by the quantum coprocessor. These discoveries were made using pen and paper and very little emulation. They wanted to see if the recently discovered effects were reflected in Google’s recent experimental studies.

Next, Skoltech’s Quantum Algorithm Lab approached the CDISE supercomputing team, led by Oleg Panarin, about the key computing resources needed to emulate Google’s quantum chips. Dr. Igor Zacharov, senior research scientist and member of the Quantum Lab, worked with several others to transform existing emulation software into a format that allows for parallel computing at Zhores. A few months later, the team created an emulation that generates data with the same statistical distribution as Google, demonstrating the range of instance densities where QAOA’s performance drops sharply. They also revealed that Google’s data is on the borderline of that range. Beyond that, current advanced technology is not enough to generate benefits.

The Skoltech team first discovered that accessibility flaws (performance constraints caused by problem constraints and varying ratios) existed in a type of problem called maximum constraint satisfiability. However, Google has considered minimizing the energy function of the graph. Since these issues belong to the same class of complexity, they gave the team conceptual hope that the issue and its subsequent impact could be related. This intuition turned out to be correct. Data is generated and the results show that the lack of accessibility creates a sort of avalanche effect that puts Google’s data at the end of this rapid transition and beyond that requires longer and longer QAOA circuits. powerful. It has been clearly shown that this would be the case.

Oleg Panarin, Head of Data and Information Services at Skoltech, commented: It has been a long and rewarding project. We worked with Quantum Labs to develop this framework. We believe this project will establish the baseline for future demonstrations of this type using Zhores. “

Igor Zacharov, Senior Researcher at Skoltech, said: “We took the existing code from Akshay Vishwanatahan, the first author of this study, and converted it into a program that runs in parallel. The data has finally emerged at an exciting time for all of us. We had the same statistics as Google. In this project, we have created a software package that can emulate a variety of advanced quantum processors up to 36 cubes and 12 layers deep. “

Akshay Vishwanatahan, PhD student at Skoltech, concludes: The internal emulation software we developed can only handle toy model cases, and at first I thought this project was an exciting challenge, but almost impossible. Fortunately, I was in an upbeat and ambitious peer group. This allows Google to track and reproduce data noiselessly. It was certainly a time of great excitement when our data matched Google data and we got a similar statistical distribution. From there, I was finally able to confirm the existence of the effect. “

See also: V. Akshay, H. Philathong, I. Zacharov, J. Biamonte, “Failures in Reachability in Quantum Approximation Optimization of Graph Problems”, August 30, 2021 quantum..
DOI: 10.22331 / q-2021-08-30-532

Russian scientists use supercomputer to probe the limits of Google’s quantum processor Source link Russian scientists use a supercomputer to probe the limits of Google’s quantum processor

Source link

About Mariel Baker

Check Also

Will Nepo’s supercomputer give him an edge over Carlsen for the world title? | Chess

IIt would count as one of the most seismic shocks in modern chess history if …