Thoughts on IBM's Zero Noise Extrapolation (ZNE) for Quantum Computing
IBM Quantum recently published a really interesting paper in Nature regarding Evidence for the utility of quantum computing before fault tolerance. Basically, a major limiting factor for quantum computing right now is the presence of noise in these systems that causes errors. Quantum systems are inherently noisy in that they tend to decohere due to environmental interactions. Much work is being done on quantum error correction algorithms to overcome these issues.
IBM's published results show evidence for getting around the noise problem by running quantum calculations through a 127-qubit quantum computer at varying non-zero noise levels, and then extrapolating out the calculation to its zero-noise counterpart by leveraging those empirical noisy data points. It's great to see that this works well for the spin system expectation values calculated in the paper.
I do have a couple of thoughts regarding the approach:
- It seems challenging to generalize ZNE in a robust way since there is an ambiguity regarding choice of extrapolation scheme. Interpolation is one thing, but high-fidelity extrapolation has always been a thorny numerical problem. The authors utilize exponential and polynomial extrapolation functions, but one wonders if a broader set of functions may be needed for a broader set of calculations and what the appropriate logic would be for the right choice.
- The authors are able to validate the ZNE approach for calculations where the exact answers are already known, but the manner of validation for novel problems where exact classical answers are unavailable is unclear.
One could imagine running subsequent experiments for other Hamiltonians and expectation values with known exact classical results, determining the appropriate extrapolation schemes, and then formulating an overall framework for ZNE by leveraging the collective insights.
Overall, it seems to me that large quantum computers with high qubit counts would be the more robust approach to error correction due to the ability to incorporate redundancy by spreading quantum information across excess qubits via entanglement. Understandably, we are still progressing toward larger machines, so it is very exciting to see positive progress being made with the current pre fault tolerant hardware!
References:
- Quantum error correction - Wikipedia
- Evidence for the utility of quantum computing before fault tolerance | Nature
- New IBM, UC Berkeley paper shows path toward useful quantum | IBM Research Blog
Comments
Post a Comment