Zoë Holmes adds expertise in quantum machine learning and an open-minded approach to the field to MARVEL’s phase 3

by Nicola Nosengo, EPFL, NCCR MARVEL

After studying philosophy of physics for her Master’s at the University of Oxford, Zoë Holmes did her PhD at Imperial College London on quantum information theory. After a three month stay at the University of Exeter, she moved to the Los Alamos National Laboratories as the Mark Kac Postdoctoral Fellow working on quantum algorithms, before joining EPFL in September 2022. 

Have you always been interested in science?

I partially conform to the gender stereotype. I liked science and maths in my mixed primary school, but then for secondary school I went to a girls’ school where science was not respected. You were considered smart if you were good at history, English, debating. Maths was kind of respected, but science…no one cared much about it. I love literature and poetry, I am not a pure scientist. But I had a physics teacher when I was 15, and I had a brilliant love/hate relationship with him. I found physics quite easy. And I was very keen and annoying, asking him a lot of questions sometimes (‘is the universe deterministic?’ was one of my favourites) but paid no attention whatsoever at others (because it was all quite straightforward). However, he would set harder exams and although they annoyed my contemporaries I realized I really enjoyed them. So, I thought, maybe I could do this. In the UK you can give up science in school at 16, and I probably would have done so if not for him. I would have probably gone into philosophy. But at 18 I was still quite enjoying physics, and I decided to study philosophy and physics in university, still thinking I was going to drop physics at some point. When the moment to start my PhD came, I was tempted by philosophy of physics, but there was not much funding for it, so I thought doing more physics was more sensible. And I’m so glad I did. I love physics research and I’ve found that a lot of the skills I learned and enjoyed thanks to the philosophy of physics are skills that I use daily as a researcher.  

Why are you a scientist? What attracts you to this profession?

I really, really enjoy it, and that’s what I look for when I am recruiting. I don’t want people who are ambitious for the sake of ambition. I want to recruit students who want to do a PhD because they just find getting to the bottom of a problem really fun!

What is the main focus of your current research?

A lot of what I am doing right now is figuring out how to make the most of near-term quantum computers. There’s a range of opinions within the community. You have the evangelical start-ups that are claiming we are going to transform everything from materials science to quantum chemistry within five years. On the other side you have those who say quantum computers are never going to work, or not without full error correction. I am somewhere in the middle. My stance is “I don’t know”. I don’t really know if near-term quantum computers are going to be useful- but it’s a fun challenge. Let’s see if we can figure out what can and what can’t work. In fact, a lot of my research is about saying “look, these things that people have proposed are not going to scale”, but I think some of these negative results are really the first steps to figure out what we might actually be able to do. 

In all honesty, the main focus of my work is on algorithm design, and a lot of the core ideas are agnostic with respect to applications. But some of the most exciting applications are definitely in the field of electronic structure calculation and quantum simulation. We are increasingly aware that most application-agnostic approaches are unscalable. In parallel, why build something completely naïve when we can build on decades of research? That is why I am looking forward to being involved in MARVEL. My pure quantum machine learning background gives me an understanding of what we know can work and can’t work in general, and I can combine that with the expertise of people who work directly on problems with a different angle.

Which publications are you most proud of?

One would probably be “Connecting ansatz expressibility to gradient magnitudes and barren plateaus”. It’s a really simple paper but it was important for the community. The idea at its core is that problem-agnostic approaches are not going to work. That’s one of my key papers on the topic of trainability of quantum systems. The other one is my very recent paper “Out-of-distribution generalization for learning quantum dynamics” which is all about showing ways in which quantum machine learning could be easier than might be expected.

We hear a lot these days about a crisis in scientific vocations, especially in fields that suffer a lot of competition from industry – such as yours. Do you see that happening? Is it becoming more difficult to recruit PhD students and postdocs?

My experience is that for PhDs it’s still fine. There are a lot of young people who want to do a PhD in quantum computing, quantum information et cetera. Postdocs yes, it’s much harder. I guess it’s competition with industry and the lack of a clear career trajectory. There’s a big dose of chance in having a proper academic career, we see a lot of brilliant postdocs not being able to get a position. I don’t see any easy solution unfortunately.

What are your hobbies?

I put a lot of time into climbing, and there too I enjoy the problem-solving part. I take it almost as seriously as science and I think it activates the same part of my brain.