Unique quantum mechanical phenomena-based technologies are expected to become commonplace soon. These may include devices that input and output data using quantum information, which requires meticulous verification due to inherent uncertainties. Verification becomes more complicated when the device is time-dependent, and the output relies on previous inputs. Time-dependent quantum devices can be checked faster by adding a memory effect that these devices already have.

Researchers used machine learning for the first time to speed up the process.

While quantum computers create headlines in the scientific press, most experts believe these machines are still in their infancy. A quantum internet, on the other hand, maybe closer to reality. This growth would provide considerable security benefits over today's internet, among other benefits. However, even this will rely on technologies that have not yet left the laboratory. While many of the principles of the devices that will enable our quantum internet have been worked out, quantum computers must overcome other engineering obstacles before the government can commercialise these devices. However, significant effort is being made to develop tools for making quantum devices.

Quoc Hoan Tran and Associate Professor of Tokyo's Graduate School of Information Science and Technology have pioneered just such a tool, which they believe will make validating the behaviour of quantum devices more efficient and precise than it is currently. Their contribution is a technique for reconstructing the operation of a time-dependent quantum device simply by studying the relationship between the quantum inputs and outputs.

"Quantum process tomography is a technique for describing a quantum system based on its inputs and outputs," Tran explained. "Numerous researchers, however, now report that their quantum systems exhibit some memory effect, in which prior ones influence current states. This quantum process tomography means that a simple examination of the input and output states cannot adequately describe the system's time-dependent nature. You could simulate the system periodically as time passes, but this would be incredibly inefficient computationally. Our objective was to embrace and exploit this memory effect, rather than using brute force to defeat it."

Tran and Nakajima developed their unique approach using machine learning and quantum reservoir computing technology. This algorithm learns patterns of inputs and outputs that change over time in a quantum system and effectively predicts how these patterns will change in the future, even in scenarios the programme has not encountered yet. Because the algorithm does not require knowledge of the inner workings of a quantum system, as a more empirical method would, but merely the inputs and outputs, the team's methodology can be simplified and generate results more quickly.

"While our programme can currently imitate a certain type of quantum system, hypothetical systems may have a wide range of processing capabilities and memory effects. Thus, the next stage of research will be to extend the capabilities of our algorithms, thereby making them more general-purpose and thus more useful, "Tran stated. "I'm thrilled by the potential of quantum machine learning approaches and the speculative devices that they could lead to."

Conclusion

Quantifying and validating the control level used to prepare a quantum state are critical problems in quantum device construction. The quantum state is defined experimentally by tomography, a process that demands a large number of resources. However, tomography for a quantum device with temporal processing has not been determined, fundamentally different from ordinary tomography. The researchers have created a practical and approximation tomography solution for this unique case by utilising a recurrent machine learning architecture. The technique is based on frequent quantum interactions between a quantum reservoir system and a stream of quantum states.

For more information, refer to the article.

https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.127.260401 Image credits: Unsplash