Scientists use machine learning to efficiently and accurately validate quantum devices

2023-12-25

In the foreseeable future, more technologies will utilize quantum mechanics. These technologies may include devices that use quantum information as input and output data, which require careful verification due to inherent uncertainty. If the output of the device depends on past inputs, validation becomes more challenging. For the first time, researchers have utilized machine learning to greatly improve the validation efficiency of time-dependent quantum devices by incorporating certain memory effects present in these systems.

Quantum computers have made headlines in scientific media, but most experts believe that these machines are still in their early stages. However, a quantum internet may be closer to what it is now. Compared to our current internet, this will provide significant security advantages, as well as other aspects. But even so, it will still rely on technologies that have not yet seen dawn outside the laboratory. Although many fundamental principles that can create our quantum internet devices may have been studied, there are still many engineering challenges to implement these products. But many studies are underway to create tools for designing quantum devices.

Quoc Hoan Tran, a postdoctoral researcher from the Graduate School of Information Science and Technology at the University of Tokyo, and Associate Professor Kohei Nakajima, have developed a tool that they believe can make verifying the behavior of quantum devices more effective and precise than currently available. Their contribution is an algorithm that can reconstruct the working principle of time-dependent quantum devices by simply learning the relationship between quantum inputs and outputs. This method is actually very common when exploring a classical physical system, but the storage of quantum information is generally challenging, which often makes it impossible to achieve.

Tran said: The technique of describing a quantum system based on input and output is called quantum process tomography. However, many researchers now report that their quantum systems exhibit some kind of memory effect, where the current state is influenced by the previous state. This means that simple checks of input and output states cannot describe the time-dependent nature of the system. You can repeatedly establish a system model after each time change, but This will be extremely inefficient in calculation. Our goal is to accept this memory effect and apply it to our strengths, rather than using brute force to overcome it.

Tran and Nakajima turned to machine learning and a technique called quantum storage computing to establish their new algorithm. This can learn the time-varying input and output patterns in quantum systems, and effectively guess how these patterns will change, even before the algorithm has witnessed them. Due to the fact that it does not require knowledge of the internal workings of quantum systems like more empirical methods, but only requires knowledge of inputs and outputs, the team's algorithm can be simpler and produce results faster.

Tran said, "Currently, our algorithm can simulate a certain quantum system, but the assumed devices may vary greatly in processing power and have different memory effects. Therefore, the next stage of research will be to expand the capabilities of our algorithm, basically making some things more universal and useful. I am excited about what quantum machine learning methods can do and the hypothetical devices they may lead to.".

New bio ink may enable 3D printing organs to achieve frozen storage