In an effort to extend the computational ability of current-generation quantum computers, IBM announced Wednesday a method called “zero-noise extrapolation” that improves the accuracy of computations by repeating a given program multiple times with varying levels of controlled noise. Together, these calculations with varied levels of noise can be used to estimate the result of a calculation in an ideal condition where no noise exists.

The effects of environmental noise on quantum computers can be quite dramatic, even in small quantities. Computations on current quantum hardware are limited by a short coherence time–the amount of useful operational time in a calculation before quantum information is lost–and circuit depth, which measures the number of sequential operations that can be performed.

SEE: Managing AI and ML in the enterprise 2019: Tech leaders expect more difficulty than previous IT projects (Tech Pro Research)

Zero-noise extrapolation is, essentially, the idea that, “if you had some way of controllably amplifying the strength of your noise, you could then extrapolate back to what your quantum computer would have been able to compute in the absence of that noise,” IBM research scientist Abhinav Kandala told TechRepublic. “You can think of this as measuring many wrongs to reconstruct the right answer. By doing that, we essentially see that we were able to achieve accuracies that would have been otherwise inaccessible to our hardware.”

The experiment, as presented in “Extending the computational reach of a noisy superconducting quantum processor,” was performed on four qubits of a five-qubit system, though Kandala notes that “there’s nothing preventing us from scaling this to larger systems.” Though refinements in manufacturing quantum systems will reduce the amount of noise that occurs naturally, this is a near-term solution for currently available Noisy Intermediate-Scale Quantum (NISQ) machines, providing 10 times improvement in accuracy in the experiment, though Kandala warns that figure will be different depending on the type of problem.

The applicability of zero-noise extrapolation will continue, as “despite all the improvements that we will have in error rates and coherence times, there will still be noise, and this noise will still affect computations that we attempt on our quantum computers,” Kandala said. “Essentially all these improvements will compound the reach of this technique.”

This method has produced observable benefits when using quantum computers, providing the ability to access longer circuit depths. “In the context of a chemistry simulation, we could prepare states which better represent the states of the molecule you’re trying to simulate, and that was giving it computation accuracy,” Kandala said. “This is what the endeavor of quantum computing is: You want to prepare states which are not so accessible to classical computation.”

IBM’s advancement brings practical use of quantum computers for businesses one step closer to reality, as enterprises are increasingly turning to quantum computers for path optimization and other logistics tasks.

This news wraps up a month of high-profile announcements for IBM’s quantum computing initiatives. On March 4, the company announced the availability of a quantum computer with a quantum volume of 16, a new high for the company, which IBM claims puts it on track to reach quantum advantage in the next decade, and doubles the performance of the systems available last year.

On March 19, IBM detailed their efforts to bring machine learning to quantum computers, creating a support-vector network on a quantum system for the first time, using the zero-noise extrapolation method.

For more on IBM Q, check out “IBM opens Q Network Hub in Tokyo to help businesses explore quantum computing,” as well as TechRepublic’s cheat sheet for quantum computing.