This article is more than 1 year old
US military employs neural networks to figure out interference in photons for quantum communications
Green Machine gets busy on photons
The US military is exploring how machine learning algorithms can clean up quantum communication technologies and help soldiers send encrypted messages to one another on the battlefield.
Tiny snippets of information can be stored as photons of light. The bits of data are encoded in the particle’s properties, like its polarization or phase. Reading the quantum information, however, can be tricky since photons tend to get distorted by environmental factors as they get transmitted.
Funded by the US Army Research Laboratory, a team of boffins led by Louisiana State University (LSU), have figured out a way to correct these distortions in order to preserve data stored in the photons using machine learning. First, the properties of single photons are captured and processed as images.
Changes in temperature and atmospheric pressure, however, can disrupt the photons and change these images. So the team built a database of imperfect images that describe the light profiles of photons that have been impacted by these distortions. By training a neural network on the original, high quality images and its degraded versions, it can learn to restore botched images.
Useful quantum computers will be impossible without error correction. Good thing these folks are working on itREAD MORE
“These networks behave as an artificial brain that is trained with multiple distorted spatial profiles of photons produced by turbulence with different conditions, for example temperature,” Omar Magaña‐Loaiza, principal investigator of the experiment described in a paper published in Advanced Quantum Technologies, and an assistant professor of physics at LSU, explained to The Register.
“The neural network translates the distorted photon profiles to numbers that can be understood by a computer. These networks rely on a series of operations that enable the implementation of image filters: these filters are like those used for image recognition.”
In short, machine learning algorithms can learn to correct these distorted images - or photon profiles - to accurately recover the quantum information stored within it.
“In classical protocols for communication, people send strings of zeros and ones, these possibilities are represented by two states, that lead to the concept of a bit of information. In the quantum context, these bits are quantum states known as quantum bits or qubits,” he added.
“In our case, we have a larger alphabet; instead of using only zeros and ones, each single photon can be prepared in more than two states. These states are known as qudits.
"In the communication protocol used in our experiment, we send strings of qudits. By doing this, we can send more than one bit of information in a single photon. If there is an error, we correct it using artificial neural networks.”
If quantum information containing an encrypted message was stored as a single photon, the neural network can make sure that they’re decoded more accurately. The US military is interested in quantum communications for a number of reasons. In theory, it’s more secure than sending data through code, which could be cracked.
“Distributed quantum information and quantum networking research are still in their early stages,” Dr Sara Gamble, program manager, Quantum Information Sciences, at the Army Research Office, told us.
“The work described in this paper is particularly exciting because it shows enormous promise for overcoming a common type of distortion that befalls quantum information during its distribution. There are, however, many hurdles still left to overcome before most real world applications can be realized.” ®