As artificial intelligence and machine learning algorithms have received attention for making accurate predictions–in everything from judging the outcome of human rights trials to predicting the winner of the Kentucky Derby to identifying cancer–another new technology has now been applied to the task of reasoning: quantum computing.
In a new paper, scientists at Cambridge Quantum Computing exhibited how quantum computing, still a nascent field, can be useful in making practical decisions. The aim, according to the head of the Quantum Machine Learning division of CQC, Mattia Fiorentini, was to show that quantum computers can handle intuitive reasoning–using inference on a probability model–which hadn’t been approached this way before.
“It’s interesting to see how a new technology can be applied to study a longstanding problem in artificial intelligence, such as reasoning,” Fiorentini said.
The results show how simulators on an IBM Q quantum computer could handle reasoning tasks including “inference on random instances of a textbook Bayesian network, inferring market regime switches in a hidden Markov model of a simulated financial time series, and a medical diagnosis task known as the “lung cancer” problem.” In order to do this, they had to “code a quantum computer, which is not that easy, and then apply theoretical work to quantum computers,” Fiorentini said.
“We are not comparing two different theoretical approaches,” he added. “We are comparing two computational approaches to implementation for the same problem.”
The results show that the quantum machine could use inference models to draw conclusions. Probabilistic inference, which means the incorporation of uncertainty into computer programming, is particularly suited to quantum computers, Fiorentini said, because “quantum models have been proven to be more expressive, easier to train under certain circumstances.”
In practical terms, this means that quantum computing can be useful to solve both scientific and engineering problems. The results are “quite flexible, surprisingly robust, and can be applied in many fields,” said Fiorentini.
For instance, he added, Bayesian networks have traditionally been used in predictive maintenance of mission-critical equipment, such as jetliners and jet engines. “You model a system, and then you perform inference on the model by asking certain questions and by figuring out if the system is stable, reliable, and robust–or is about to break down–so you can intervene,” Fiorentini said. “And which part is signaling the stress more strongly?”
SEE: Quantum computing: An insider’s guide (free PDF) (TechRepublic)
Medical diagnostics is another field that can benefit from these results. Although it can’t be exactly applied from the results of this study, “continuing in this direction, some of these techniques are applied to drug discovery,” Fiorentini noted. “In particular, to tailor precision medicine, or treatment based on genomic information. People can be cured with a high level of efficacy if you are able to map a specific drug to a specific person, tailoring it to their genome,” he said.
Establishing a causal relationship is the ultimate goal, and still more advanced than the methods used here. Still, “this Bayesian network type and performing inference are some of the tools that are part of the toolbox,” Fiorentini said.
For instance, doctors could use this model to identify symptoms. “The first level of hidden information you might want to uncover is: What’s the cause of the symptom? Is it a common cold, or is it cancer? If it’s cancer, we can go deeper: What’s the impact of smoking?” Tackling these kinds of problems could be a future goal of the research.
Machine learning researchers, as well as quantum software developers, could draw from the research, in particular.
The widespread use of quantum computing is still a ways off. Still, “we are happy about the results. the quantum computing succeeded–and the build of this technology is still in the early stages,” Fiorentini said.
- DevOps: A cheat sheet (TechRepublic)
- Inside UPS: The logistics company’s never-ending digital transformation (free PDF) (TechRepublic)
- Microsoft Build 2020 Highlights (TechRepublic Premium)
- Technology that changed us: The 1970s, from Pong to Apollo (ZDNet)
- Tom Merritt’s Top 5 series (TechRepublic on Flipboard)