Combining two of the most recent technical advances—quantum computing and machine learning—is one of the hottest academic subjects right now.
An experimental investigation found that small-scale quantum computers can already improve the performance of machine learning algorithms.
An international team of researchers from the University of Vienna demonstrated this using a photonic quantum processor. The study, published in Nature Photonics, demonstrates exciting new uses for optical quantum computers.
Recent scientific advancements have impacted the evolution of future technology. On the one hand, machine learning and artificial intelligence have already transformed our daily lives, from household activities to scientific research. On the other side, quantum computing has arisen as a new computational paradigm.
A new study area has emerged as a result of the confluence of these promising two fields: Quantum Machine Learning. This topic seeks to identify possible improvements in algorithm speed, efficiency, and accuracy when performed on quantum systems. It remains an open task to attain such an advantage on current-generation quantum computers. However achieving such an advantage on contemporary quantum computers remains an open question.
This is where an international team of researchers went on to create a unique experiment, which was carried out by experts from the University of Vienna.
The setup includes a quantum photonic circuit constructed at the Politecnico di Milano (Italy) that runs a machine learning algorithm initially suggested by Quantinuum researchers in the UK. The objective was to categorize data points with a photonic quantum computer and isolate the contribution of quantum phenomena in order to better understand the advantage over conventional computers.
The experiment demonstrated that even small-scale quantum computers may outperform traditional algorithms.
They discovered that for particular jobs, their algorithm they use makes less mistakes than its traditional version, according to research leader Philip Walther of the University of Vienna.
This means that current quantum computers may operate well without requiring pushing beyond cutting-edge technology, according to Zhenghao Yin, first author of the Nature Photonics study.
Another fascinating element of the new research is that photonic platforms can use less energy than traditional computers. This might be critical in the future, given that machine learning algorithms are becoming unfeasible owing to excessive energy needs, says co-author Iris Agresti.
The researchers’ findings have implications for both quantum computation (by identifying jobs that benefit from quantum effects) and standard computing.
Indeed, new algorithms inspired by quantum architectures might be developed, resulting in improved performance and lower energy usage.
