IBM Corp. researchers reckon they’ve come up with new algorithms capable of enabling advanced machine learning on quantum computers.

In a paper published today on arXiv, a repository for non-peer-reviewed academic papers, IBM’s research team describes how it has created a “quantum algorithm” that enables such computers to perform “feature mapping” at a scale that goes far beyond what classical computers can do.

Quantum computing takes advantage of the strange ability of subatomic particles to exist in more than one state at any time. Thanks to the way the tiniest of particles behave, operations can be done much more quickly and use less energy than classical computers.

In classical computing, a bit is a single piece of information that can exist in two states – 1 or 0. But quantum computing uses quantum bits, or “qubits,” that can store much more information than just 1 or 0, because they can exist in any superposition of these values.

Feature mapping relates to a process of disassembling information in order to get access to “finer-grain aspects” of that data, IBM’s research team explained. Traditional machine learning algorithms can already do this to an extent, for example by taking the pixels of an image and placing them in a grid based on each one’s color value. The algorithms then map these values in a nonlinear fashion to a high-dimensional space, essentially breaking down the data according to its most useful features.

With IBM’s new quantum algorithms, however, it becomes possible to separate aspects and features of that data even to an even greater degree, the researchers said. This is important, because the more precisely data can be classified, the more efficient machine learning systems will perform.

“The goal is to use quantum computers to create new classifiers that generate more sophisticated data maps,” IBM’s research team said. “In doing that, researchers will be able to develop more effective AI that can, for example, identify patterns in data that are invisible to classical computers.”

IBM’s researchers note that the new algorithms haven’t yet achieved “quantum advantage,” which is the point at which quantum computers surpass the performance of classical machines. That’s mainly because quantum computers are still in their infancy, limited by current hardware capabilities, IBM said.

“Our research doesn’t yet demonstrate quantum advantage because we minimized the scope of the problem based on our current hardware capabilities, using only two qubits of quantum computing capacity, which can be simulated on a classical computer,” IBM’s researchers said.

Nonetheless, IBM’s work once again demonstrates how quantum computing promises to operate next-generation applications far better than any compute infrastructure that’s currently available, Holger Mueller, an analyst with Constellation Research Inc., told SiliconANGLE.

“IBM has shown how selected machine learning algorithms like feature mapping run better on quantum computers than anything else,” Mueller said. “Feature-calling algorithms are ideally suited for quantum computing.”

IBM said its new algorithms will be made available to everyone via its Qiskit Aqua open source library for developers, researchers and other experts.

Photo: Oraelius/Flickr