Researchers have showcased that a very small quantum setup, comprising merely 9 atoms, can outperform substantial classical machine learning algorithms when applied to time-series forecasting challenges. Specifically, this was demonstrated in a practical application involving multi-day temperature prediction.
The paper’s originators maintain this marks the initial experimental verification where quantum machine learning exhibited superiority over extensive classical models in a real-world forecasting scenario. Rejecting the conventional strategy of simply increasing scale and complexity, the researchers adopted an alternative methodology: input data was fed into the system, which then spontaneously evolved, with the outcome being extracted at the end—bypassing meticulous control over every single computational step. The scientists managed the atomic manipulation by employing nuclear magnetic resonance (NMR) techniques.
Reportedly, even as the corresponding classical system was scaled up to encompass thousands of processing units, the nine-atom quantum apparatus consistently achieved greater accuracy in making predictions looking further into the future.
The team emphasizes that this does not yet represent a fully universal quantum computer; the system’s scale is restricted, and it was validated only on specific categories of problems. Nevertheless, this work clearly illustrates that tangible benefits can already be harvested from diminutive quantum architectures.
About the Author
