Rene: Hi! Welcome to QuBites, you’re bite sized pieces of quantum computing. My name is Rene from Valorem Reply and today we’re going to talk about quantum machine learning. And for this I’m honored to have a special guest Johannes Oberreuter. Hi Johannes and welcome to the show! How are you today?

Johannes: Hi Rene, thank you! I’m very good. I’m excited, how are you doing?

Rene: Perfect! And since you are from Bavaria - one of the states in Germany that probably drinks the most beer, and I’m from Saxony - another state that loves to drink beer, and it’s Friday afternoon, we said let’s have a beer and talk about quantum machine learning. So cheers my friend!

Johannes: Cheers, here’s to quantum!

Rene: Yeah! Alright, so my first question is well, tell us a little about yourself and your background as it relates to quantum computing.

Johannes: Sure. I actually did study Physics and went on to do a PhD in Quantum Gravity and then I switched to dealing in my post doc with Quantum Condensed Matter Systems and their dynamics. And after that I started to work as a Data Scientist, three years ago, for Machine Learning Reply. And for the past year I’ve been co-leading the Reply Quantum Computing Community of Practice.

Rene: Wow that’s impressive actually! So again, a PhD in physics and all that background, so we could go very deep but don’t worry folks we’re actually talking about applied quantum computing. And in fact, tell me a little bit about what our topic is all about today, what is quantum machine learning?

Johannes: Well already in it’s name it has these two components, right? Quantum and machine learning. So machine learning has been around for a couple of years now and essentially it is trying to teach computers how to evaluate large amounts of data which would either be too complicated for a human to evaluate because the information you’re looking for is somewhere deep inside the data, monolinearly hidden, or it is just too much data. So, machine learning has been performing really well, it has created a lot of business, a lot of new opportunities and some ongoing field of research of course. And now the other part, the quantum, indicates that we now try to get advantages for machine learning from quantum computing. So why would you think that there might be an advantage?

Well machine learning needs a lot of resources, it’s very computationally [intensive] and quantum computing promises to speed up certain processes. So you might have been talking about breaking of encryption algorithms and the breaking here is really only possible because a quantum computer can perform very specific tasks much more efficient than a classical computer - conventional hardware we sometimes also call classical computers. So that is why you can hope that a quantum computer can be better. And indeed this is something you partially see for certain machine learning tasks, there is indeed a speed up, a quantum speed up. That is also what Google calls quantum supremacy but there is a second part which I find even more intriguing and really fascinating. Which Is that the use of quantum machine learning algorithms allows you to use certain procedures which make the result of the machine learning more accurate. It really improves the result we’re getting. Not only the speed but also the result. That is something I would not have thought would be possible before we started working in this field, I’m really fascinated by this.

Rene: That’s impressive, so it’s not just faster it’s more precise. Because typically in classical computing these are opposite things, right? You can either get it faster and less precise but getting both, I mean that’s just gold right? So, let’s talk a little bit about the impact we’re already seeing today with quantum machine learning. And maybe you can share some examples where you are seeing impact with quantum machine learning and especially in which kind of key areas and key industries you see those.

Johannes: Yeah, thank you. I mean to be honest, the whole thing - quantum machine learning - is still quite research’y and experimental. That has also to do with the state of the hardware. You have probably already told your audience and discussed with them in the previous episodes that quantum hardware is not as advanced as classical hardware is. So you can access a quantum computer via the cloud however, via the cloud you can also get GPUs. And GPUs are super-efficient at training neural networks and deep learning and they are cheap and they are mass produced and that is certainly not true for quantum computers. So, this is not really yet production ready hardware. So, in that sense that’s a hard competition for quantum machine learning to go against these algorithms. So, at the moment it is quite experimental, we are looking into what we can do. But especially in the research area for instance with particle accelerators, like the LHC (Large Hardron Collider) that discovered the Higgs [boson] particle a couple of years ago, they are already trying these algorithms and they do see that the quantum classification algorithms are a bit better or at least as good as the algorithms they have been using thus far. And that is super exciting because usually those researchers are pretty good at what they do so if they can improve this, it means that is has a lot of potential.

Rene: So basically, they are applying quantum machine learning to solve more quantum mechanical challenges. That’s just exceptional right?

Johannes: Exactly. Isn’t that funny?

Rene: Yeah! You wanted to share some more key industries I think.

Johannes: Yeah so that model of success is to try it ourselves and implement and of course all the benchmarks. Because as a consulting company, of course, we can not only sell enthusiasm to our clients we also need to make sure that what we are doing with them is an improvement on what they have so far and that it’s safe to use. So, we tried this actually, several algorithms - a classification algorithm for instance, on a client data set and we saw that the performance was at least as good as the conventional algorithms. And secondly, of course also the implementation will improve along with the hardware. So, I really see a lot of potential.

So where do we see this potential? I think the first applications will be in the financial industry. There is a lot of interest and there is also really the need for speed. If you think about high frequency training and these really high dimensional data sets that they are dealing with, like thousands of assets. That is something which you really, at the moment, cannot really optimize and analyze with conventional computers so there’s definitely a big possibility for quantum computers. We see a lot of literature also popping up in this direction from improving your training to predicting the next financial crash. So, there is a lot of interesting development we can expect.

Rene: Well, that was awesome, thanks for sharing all your insights. And we’re already at the end of the show. Thank you so much Johannes for joining us and sharing your insights again very much appreciated.

Johannes: It was a real pleasure. Thanks, Rene, for inviting me.

Rene: Cheers my friend and enjoy your Friday afternoon!

Johannes: Enjoy the weekend!

Rene: Bye, bye.

Johannes: Bye.

Rene: Thanks everyone for joining us for another episode of QuBites. Watch our blog, follow our social media channels to hear all about the next episodes. Take care, be safe and see you soon, bye bye.