We are getting closer to a new era of quantum computing as EPFL researchers demonstrate that even a few straightforward examples are sufficient for a quantum machine-learning model, the “quantum neural networks,” to learn and predict the behavior of quantum systems.
Imagine a world in which computers can decipher the enigmas of quantum mechanics, allowing us to conduct incomparably accurate simulations of the complex dynamics of molecules or examine the behavior of complicated materials.
We are now a step closer to that realization thanks to a ground-breaking study conducted by Professor Zoe Holmes and her group at EPFL. Together with scientists from Los Alamos National Laboratory, the Free University of Berlin, and Caltech, they have developed a novel technique for instructing a quantum computer on how to comprehend and anticipate the behavior of quantum systems. Nature Communications has published the study.
Quantum neural networks (QNNs)
The team focused on “quantum neural networks” (QNNs), a class of machine-learning models created to learn and process data using ideas from quantum physics to resemble the behavior of quantum systems.
QNNs are composed of interconnected nodes, or “neurons,” that carry out calculations, just like the neural networks used in artificial intelligence. The distinction is that QNNs can handle and manipulate quantum information since their neurons function according to the laws of quantum physics.
“Normally, we need a lot of examples when teaching a computer something,” claims Holmes. However, in this study, we demonstrate that even when dealing with entangled states, which are more complex and difficult to understand, the computer can learn how a quantum system operates with just a few basic examples dubbed “product states.”
Product states
The term “product states” employed by researchers refers to a quantum mechanics idea that describes a particular kind of state for a quantum system. For instance, if a quantum system consists of two electrons, the product state is created by first considering the states of each electron separately, and then combining them.
In order to study and understand the behavior of quantum systems before moving on to more complex and entangled states, where the particles are correlated and cannot be described independently, product states are frequently used as a starting point in quantum computations and measurements.
Better quantum computers ahead
The researchers showed that computers can efficiently understand the complicated dynamics of entangled quantum systems by training QNNs using just a handful of these straightforward examples.
According to Holmes, “This means that [we] might be able to learn about and understand quantum systems using smaller, simpler computers, like the near-term intermediary scale [NISQ] computers we’re likely to have in the coming years, instead of needing large and complex ones, which may be decades away.”
The research also opens up new avenues for employing quantum computers to tackle challenging issues like researching intricate new materials or mimicking molecular activity.
Finally, by enabling the development of shorter and more error-resistant programs, the technique enhances the performance of quantum computers. We can simplify the programming of quantum computers, resulting in increased efficiency and dependability, by understanding how quantum systems operate. Making programs for quantum computers more concise and error-free would improve their performance, claims Holmes.