Researchers at the Autonomous University of Barcelona have developed a new protocol that classifies and orders quantum data according to the state in which they were prepared. This procedure, which offers greater efficiency than the equivalent classical algorithm, is a new step towards machine learning in quantum information networks.
The technologies based on computing and quantum information promise unprecedented applications, such as unconditionally secure communications, ultra-precise sensors, and quantum computers capable of solving certain problems with unattainable efficiency in classic computers.
Recently, the construction of a large network of quantum information exchange devices has been raised, where connections are established through quantum channels and the data itself is quantum systems that flow through the network, thus laying the foundations of a future “quantum internet “
With the design of these quantum information networks, new theoretical challenges appear, since it is necessary, for example, to establish optimized information treatment protocols to work with quantum data automatically, in the same way that current communication networks automatically manage information.
Now, researchers from the Autonomous University of Barcelona (UAB) have faced for the first time one of these challenges: the problem of ordering data from a network of quantum systems according to the state in which they were prepared. The researchers, who publish their study in the journal Physical Review X, have devised an optimal procedure that allows grouping those quantum systems that have been prepared identically.
This protocol shows a natural connection with a typical case of machine learning: grouping data samples that have been generated according to the same probability distribution.
The problem is similar to what a classic computer would have to solve to discern where different sounds come from simultaneously captured by a microphone in the street. The computer could recognize patterns and discern a conversation about traffic noise or sound coming from a street musician’s guitar.
However, unlike what happens with sound waves, identifying patterns among quantum data is a challenge, since the observation of these data only provides partial information about them and, in addition, irreversibly degrades them.
The physicists of the UAB have also been able to compare the performance of the classical and quantum protocols. According to the researchers, the new protocol far exceeds the classical strategy, especially when the size of the data is large.
The proposal represents a new step towards quantum information networks, since it establishes a solid theoretical framework on what is physically possible in the field of automatic classification and distribution of this type of information.