Vector researchers use machine learning to build better quantum computers

September 29, 2021

Blog Insights Machine Learning

By Ian Gormely

September 29, 2021

Vector researchers are applying machine learning techniques to build better quantum computers, which use qubits instead of bits. Quantum computers are generally faster and have more memory than the “classical” computers with which most people are familiar, making them ideal for problems related to encryption, data analysis, and optimization. 

Faculty Member Juan Felipe Carrasquilla co-authored the paper Protocol Discovery for the Quantum Control of Majoranas by Differentiable Programming and Natural Evolution Strategies,” with Luuk Coopmans, Di Luo, Graham Kells, and Bryan K. Clark, which was published in PRX Quantum in June. “It details a more efficient way to manipulate qubits based on a type of particle called Majorana zero modes using machine learning, which was never done before,” says Carrasquilla who is also an adjunct assistant professor at the University of Waterloo (UW). UW and the Vector Institute are key partners, sharing many Faculty Members, Faculty Affiliates, Postdocs and students. 

In applying ML techniques to the field, the team looked at one of quantum’s most basic optimization questions — how can quantum information, specifically Majorana zero modes, move from point A to point B in the quickest way possible — and gamified it using techniques similar to reinforcement learning. “The game starts, and initially, you’re not very good,” he explains. “You try again. and then you start figuring out what the qubit’s movement strategies could be.” 

Logic would suggest that moving something at a constant speed in a straight line would make the most sense, but this was not the case. Instead, the Majorana zero mode encoding the quantum information needs to jump around really quickly, arriving at its destination in a manner that Carrasquilla likens to hopping in and out of a moving bus. 

Carrasquilla believes that there is potential to apply these techniques to a number of different quantum-related problems. The paper has already been cited by a number of colleagues, including fellow Vector Faculty Member Alán Aspuru-Guzik who used some of the techniques Carrasquilla and his team laid out to optimize quantum circuits

But its biggest impact so far has been in a realm he had never considered: the microscopic refrigerators used to cool electronic devices and fibre-optic networks. “I was contacted about using this strategy to do something which is completely different from what I had considered,” says Carrasquilla. “You can never tell where your work is going to end up.” 

Click here for more information about Juan Felipe Carrasquilla and his work. 


Vector Faculty Member Gautam Kamath

Vector researcher Gautam Kamath breaks down the latest developments in robustness and privacy

Large Language Models

Standardized protocols are key to the responsible deployment of language models

Machine Learning

The known unknowns: Vector researcher Geoff Pleiss digs deep into uncertainty to make ML models more accurate