Researchers from Russia’s Skolkovo Institute of Technology (Skoltech) have developed a new method to speed up the calculation of quantum interactions,media reported. They complete the entire process on quantum neural networks, rather than storing/computing quantum information on classic computers through classical algorithms. Unpredictability is an inherent problem in the modeling of quantum scale interactions. Because there are few theoretical models that can predict the results of complex interactions, scientists rely on sampling techniques.
They calculate the same thing over and over again, add a degree of randomness, and finally assess the overall picture. Although this produces effective results, it requires a lot of computing power.
However, Skoltech combines some theoretical methods in the development of quantum computing, replacing the randomness of sampling methods with the special characteristics of quantum computers. Their method uses an algorithm called a variable quantum feature solver to create a quantum description of the starting position of all interacting objects/forces. Then add some additional information to the location of the classic neural network to estimate the type of interaction. After that, quantum neural networks (which are still theoretical) calculate interactions and search for patterns in the output.
Numerical tests found that the researchers’ methods yielded moderate accuracy: most of the voting quantum classifiers they used were trained as test networks to identify phase snares in the horizontal field Ising model with 99% accuracy and 94% accuracy in identifying phases in the XXZ model. However, quantum computers that work in this way have not yet been designed, so it’s too early to celebrate.