Physicists from the Swiss Federal Institute of Technology Lausanne (EPFL) and Columbia University have presented an approach to simulating a quantum algorithm using a conventional computer. The new approach uses a classic machine learning algorithm that mimics the behavior of quantum computers in the near future.
In a paper published in the journal Nature Quantum Information, EPFL professor Giuseppe Carleo and Columbia University graduate student Matija Medvidovic have found a way to perform complex quantum computing algorithms on traditional computers instead of quantum ones.
The “quantum software” known as the Quantum Approximate Optimization Algorithm (QAOA) is used to solve classical optimization problems in mathematics. Basically, it is a way of choosing the best solution to a problem from a variety of possible solutions. There is a lot of interest in understanding what problems can be effectively solved by a quantum computer, and QAOA is one of the most visible candidates for this, ”Carleo explained.
QAOA has many supporters, including Google, which is betting on quantum technology and computing in the near future: in 2019 they created Sycamore, a 53-qubit quantum processor, and used it to perform a task they estimate , a modern classical supercomputer would take about 10 thousand years. Sycamore completed the same task in 200 seconds.
Using conventional computers, scientists have developed a method that can roughly mimic the behavior of a special class of algorithms known as variational quantum algorithms, which are ways to determine the lowest energy state, or “ground state” of a quantum system. QAOA is one of the important examples of such a family of quantum algorithms, which, according to researchers, are among the most promising candidates for “quantum advantage” in computers in the near future.
The approach is based on the idea that modern machine learning tools can be used to train and emulate the inner workings of a quantum computer. The key tool for these simulations is the Neural Network Quantum States, an artificial neural network that Carleo developed in 2016 with Mathias Troyer and is now being used for the first time in QAOA simulation.