The July 2022 issue of IEEE Spectrum is here!

Close bar

Spooky Action Could Help Boost Quantum Machine Learning

Mysterious quantum links could help lead to exponential scale-up

3 min read
A drawing of two laptops. The left laptop, labeled classical data, shows 1s and 0s on its screen, and then interlinked lines, and finally an identified cat. The right laptop, labeled entangled data, shows the same but the third screen identifies the cat as alive.

Both conventional and quantum machine-learning approaches are hampered by the “no free lunch” theorem, which suggests that ultimately any approach is hampered by the amount of data (or entanglement) available.

Los Alamos National Laboratory

Machine learning, which now powers speech recognition, computer vision, and more, could prove even more powerful when run on quantum computers. Now scientists find the strange quantum phenomenon known as entanglement, which Einstein dubbed “spooky action at a distance,” might help remove a major potential roadblock to implementing quantum machine learning, a new study finds.

Quantum computers can theoretically prove more powerful than any conventional computer on a number of tasks, such as finding a number’s prime factors—the mathematical foundation of the modern encryption currently protecting banking and other secure data. The more components known as qubits that are linked together in a quantum computer through entanglement—wherein multiple particles can influence each other instantaneously regardless of how far apart they are—the greater its computational power can grow, in an exponential fashion.

Scientists are still researching the specific problems for which quantum computing might have an advantage over classical computing. Recently, they have begun exploring whether quantum computing might help boost machine learning, the field of AI that investigates algorithms that improve automatically through experience.

One potential application of quantum machine learning is simulating quantum systems—for instance, chemical reactions that might yield insights leading to next-generation batteries or new drugs. This might entail creating models of the molecules of interest, having them interact, and using experiments of how the actual compounds interact as training data to help improve the models.

A potential major stumbling block that quantum machine learning may face is the so-called “no free lunch” theorem. The theorem suggests any machine learning algorithm is as good as, but no better than, any other when their performance is averaged over many problems and sets of training data.

A consequence of the no-free-lunch theorem is that a machine-learning algorithm’s average performance depends on how much data it has, suggesting the amount of data ultimately limits machine learning’s performance. This raised the possibility that in order to model a quantum system, for example, the amount of training data that a quantum computer might need would grow exponentially as the modeled system became larger. This could potentially eliminate the edge that quantum computing could have over classical computing.

Now scientists have discovered a way to eliminate this exponential overhead using a newfound quantum version of the no-free-lunch theorem. Their findings, verified using quantum-hardware startup Rigetti’s Aspen-4 quantum computer, suggest that adding more entanglement to quantum machine learning can lead to exponential scale-up.

Specifically, the researchers suggested entangling additional qubits with the quantum system that a quantum computer aims to model. This extra set of “ancilla” qubits can help the quantum machine-learning circuit interact with many quantum states in the training data at the same time. As such, a quantum machine learning circuit may experience a speedup even with relatively few ancillas.

“Trading entanglement for training states could give huge advantages for training certain types of quantum systems,” says study coauthor Andrew Sornborger, a physicist at Los Alamos National Laboratory, in New Mexico.

Sornborger cautions that it can prove extremely difficult entangling ancilla qubits with the quantum systems used in the experiments needed to supply training data. Still, “as long as it is not exponentially difficult in some sense to create entanglement, then we stand to benefit,” he says.

One potential futuristic application of this work is what the researchers call “black box uploading.” “Wouldn’t it be cool to be able to learn a model of a quantum experiment, then study it on a quantum computer without having to do repeated experiments?” Sornborger says.

For example, if the atom smashers at CERN, the largest particle physics lab in the world, entangled the protons being collided together with the detectors used to analyze them and with an extraordinarily powerful quantum computer (on the order of a billion billion qubits), scientists could have a way to directly analyze the Standard Model, currently the best explanation for how all the known elementary particles behave.

“This is the sort of possibility that we’ve begun to contemplate in the quantum machine-learning context,” Sornborger says.

The scientists detailed their findings 18 February in the journal Physical Review Letters.

The Conversation (0)
Twenty people crowd into a cubicle, the man in the center seated holding a silicon wafer full of chips

Intel's million-transistor chip development team

In San Francisco on Feb. 27, 1989, Intel Corp., Santa Clara, Calif., startled the world of high technology by presenting the first ever 1-million-transistor microprocessor, which was also the company’s first such chip to use a reduced instruction set.

The number of transistors alone marks a huge leap upward: Intel’s previous microprocessor, the 80386, has only 275,000 of them. But this long-deferred move into the booming market in reduced-instruction-set computing (RISC) was more of a shock, in part because it broke with Intel’s tradition of compatibility with earlier processors—and not least because after three well-guarded years in development the chip came as a complete surprise. Now designated the i860, it entered development in 1986 about the same time as the 80486, the yet-to-be-introduced successor to Intel’s highly regarded 80286 and 80386. The two chips have about the same area and use the same 1-micrometer CMOS technology then under development at the company’s systems production and manufacturing plant in Hillsboro, Ore. But with the i860, then code-named the N10, the company planned a revolution.

Keep Reading ↓Show less
{"imageShortcodeIds":[]}