Quantum computing, the CNR signs the model that surpasses Google

The Italian researchers have figured out how to build a quantum logic gate starting from those actually implementable on a specific hardware. Challenge won thanks to artificial intelligence and deep learning 11 Aug 2021 Patrizia Licata Journalist

In research on quantum computing, Italy is confirmed more than ever as a protagonist with the result achieved by a team of researchers coordinated by Enrico Prati of the Institute of Photonics and Nanotechnologies of the National Research Council (Cnr-Ifn) in Milan. Thanks to artificial intelligence, the group of scientists has developed a quantum compiler to program an algorithm on any quantum computer based on logic gates. The result, obtained from the collaboration with Matteo Paris of the State University of Milan and with Marcello Restelli of the Milan Polytechnic, was published in the journal of Nature "Communications Physics". Quantum computer hardware is complex and offers only some fundamental logic gates, while an algorithm may require operations based on logic gates that are not among those available. But is there always a way to build a quantum logic gate starting from those actually implementable on a specific hardware? Until now, the theory has claimed yes, but without saying how to proceed. Index of topics • The "turning point" on quantum logic gates • The Italian patent "exceeds that of Google" The "turning point" on quantum logic gates "Similarly to conventional computers, in which the bits are subjected to calculations through the logic gates, even in quantum computers it is necessary to use quantum logic gates, which, however, must be programmed by a sort of operating system that knows which operations are achievable", he explains Enrico Prati. "However, there are many different versions of hardware that provide different achievable operations, like a small deck of playing cards to choose from." There is a theory according to which there is always a way to build a quantum logic gate starting from those that can actually be implemented on a specific hardware, but how to proceed? "It's a bit like looking for a way to start from Rome Fiumicino and arrive in all the cities of the world, being able to travel only exactly 15,000 km along two precise directions, along meridians or parallels, and having all the fuel available. wants. To get to Rome Ciampino it will be necessary to travel around the world many times before landing and the challenge was to solve the problem by relying on the abstraction capacity of artificial intelligence ", says Prati. The Italian patent "surpasses that of Google" The challenge was entrusted to Lorenzo Moro, at the time graduating in Physics at the University of Milan and now a PhD student at the Milan Polytechnic. "We asked the artificial intelligence to find the order to play the 5-6 cards available, even with sequences lasting hundreds of plays, choosing the right ones one by one to form the entire sequence", explains Moro. "After a training phase, which lasts from a few hours to a couple of days, the artificial intelligence learns to build the sequence for each quantum logic gate starting from the available operations, but taking a few milliseconds". The research was also patented. "Our model actually surpasses a similar patent from Google that uses artificial intelligence after training but for only one logic gate, then a new training is required. Instead, we have identified how to build all the quantum logic gates with a single training, after which the solution for any logic gate can be immediately recalled, with the so-called deep learning ”, concludes Prati. "Quantum computers promise to solve computational problems much faster than with currently existing hardware, and quantum compilers are a key element for their efficient control."