James Le

View Original

Datacast Episode 56: Apprehending Quantum Computation with Alba Cervera-Lierta

James Le

The 56th episode of Datacast is my interview with Alba Cervera-Lierta— a Postdoctoral fellow at the Matterlab group at the University of Toronto.

We had a wide-ranging conversation that covers her foray into studying physics; her research in Bell Inequalities, multipartite entanglement, and entanglement in fundamental interactions; her current work that intersects quantum computation and machine learning; her education and public outreach activities to spread awareness about quantum technologies, and much more.

See this content in the original post

Listen to the show on (1) Spotify, (2) Apple Podcasts, (3) Google Podcasts, (4) Stitcher, (5) iHeart Radio, (6) Breaker, and (7) RadioPublic

Key Takeaways

Here are highlights from my conversation with Alba:

On Studying Physics at The University of Barcelona

Source: https://www.ibm.com/blogs/research/2018/06/teach-qiskit-winner/

Since I was a child, I have always been interested in science, thanks to my father. Then, I had an outstanding physics professor in my high school who encouraged me to choose physics among other scientific inquiries.

My background is in quantum physics in specific and in quantum information in general. I am interested in fundamental questions that make quantum physics different from classical physics. Recently, I became interested in the practical side of quantum computation too.

I was the first generation at the University of Barcelona to study quantum information. I also took courses in particle physics, which was the focus of my M.S. there.

On Bell Inequalities

The Bell theorem states that quantum physics can be described with a local hidden variable theory. This was a revolutionary result in the 60s by John Bell because it proves that quantum entangled particle correlation can be described using classical physics. At the time, people suspect that there must be some intriguing properties in those particles, which quantum mechanics does not tell fully.

The Bell inequalities experiments say that by entangling two particles and separating them into a huge distance, we can measure the correlation between some physical observables of these two particles and deduce some expressions. If we can explain quantum mechanics from a classical point of view, there is a maximum bound of these inequalities. Bell proofs show that these bounds are violated theoretically if quantum mechanics exists, and the maximum value is greater than 2. This was proven experimentally 20 years later, meaning that classical physics can explain quantum phenomena.

Since then, many people have discovered new Bell inequalities that involve different states and degrees of freedom. Our work proposed new Bell inequalities that involve qutrits, a type of quantum state.

On Maximal Entanglement in High-Energy Physics

We studied the structure of fundamental theories, in particular quantum electro-dynamics — the physics behind interactions between light and matter. We examined how these types of interaction can generate maximally entangled states. We recovered known results from the symmetries argument by imposing entanglement. Nature as such, maximal entanglement exists and can be generated.

Later on, we also extended these results to another fundamental interaction theory. It was very cool to study fundamental interactions from a quantum perspective and see how entanglement is at the core of such interactions.

On Spin Chains and Hyperdeterminant

Spin chains are the physical models used to explain and study magnetism at the fundamental level. Highly-sophisticated spin chains can explain properties like superconductivity, for instance. They are essential in condensed matter physics and new materials.

Quantum phase transitions occur due to quantum fluctuation instead of the typical temperature. It has been proven that the critical points of the entanglement scale logarithmically with the size of the spin chains. We decided to study these entanglements using hyper-determinant, which can measure entanglements for every spin or every party.

On Theory vs. Experiment

Having a quantum experiment implies that you need quantum particles (a tiny field), and you need to isolate them from their environments (an arduous task). You also need to find in nature quantum mechanical systems that can be used to test your ideas.

Theoretically, you can propose many things. But experimentally, you need to find the physical systems that can represent the properties in your theories.

On The Quantic Joint Effort

Source: http://quantic.bsc.es/about/

My former advisor at the University of Barcelona, José Ignacio Latorre, decided to join forces with the Barcelona Supercomputing Center — where there were researchers with quantum experience and, more importantly, a super-computer that can be used to test and benchmark quantum algorithms.

Some researchers in the Barcelona Supercomputing Center are physicists and computer scientists with an interest in complexity theory. Quantum information is a new part of this theory, so it’s also attractive for them.

From the quantum computing point of view, we also need to know the limitations of classical computation in the first place.

On The Quantum Symphony

Source: https://albacl.github.io/post/quantum_symphony/

A symphony is a good analogy to explain quantum mechanics because it maps the physics with the music.

  • Quantum states are described by the wave function, which tells you the probability of observing some states in a particular place.

  • Each quantum state is a tone of a particular note. A combination of states is a chord, which can carry more than one tone.

  • Similarly, we can have quantum states carrying one tone or more than one tone (super-position).

  • A quantum symphony plays with different quantum states, generates different tones/chords, and outputs the final computation.

There are also different quantum instruments.

  • Some instruments are represented by pure theoreticians (people like me) who write the musical sheet (coming up with the theories).

  • Other instruments are represented by companies and research centers that play the melody (building the devices).

On Winning IBM’s QISKIT Challenge

I was running my exact ising model simulation algorithms on IBM computers. When I knew of this IBM challenge, I decided to write a tutorial and submit it to the contest. I wasn’t expecting to win, to be honest, as there were more amazing and sophisticated tutorials in submission. I believe that my tutorial won because I was able to show the performance of a quantum simulation algorithm.

Thanks to this achievement, I gave many interviews in Spanish newspapers, probably because many people discovered that quantum computers exist. I gained visibility both in academia and in the general public who are interested in this technology.

On Quantum vs. Classical Physics

Entanglement does not exist in classical physics. Entanglement is the quantum correlation. If you don’t have a high quantum correlation, you can simulate everything with a classical computer. On the other hand, if you have a high correlation, then there are some properties that you can exploit. These are used in algorithms such as Grover Search and Shor’s Factorization, which are quantum algorithms with exponential speedups.

The motivation behind building a quantum computer is to manage the entanglement that appears not only in quantum algorithms but also in the simulation of quantum phenomena (like spin chains or quantum phase transition).

On Quantum Classifier

We explored the universal approximation function theorem in quantum neural networks (also called variational quantum neural networks). We proposed the smallest possible quantum model capable of being a universal approximation function. Each neuron in the classifier is a quantum gate that encodes the data and trainable/optimizable parameters.

Source: https://arxiv.org/pdf/1907.02085v2.pdf

Our paper showed that we only need one qubit, a minimal quantum information unit, to perform a quantum classification task. Our goal was not to prove a quantum advantage but to prove that a quantum neural network can achieve the same thing as a classical neural network can even with one qubit.

Quantum computing is still in a very early stage. We did not benchmark our quantum classifier with any benchmark against state-of-the-art (classical) neural models.

  • In fact, we started from the first principle: How to define a neuron? How to put together neurons? How to train these neurons?

  • Moreover, we examined the mathematical theorem of classical neural networks and attempted to find the quantum counterpart.

On Being a Postdoctoral Fellow at The Matter Lab in the University of Toronto

I am thrilled to be a part of the Matter Lab group. We are a multi-disciplinary group working in quantum computing, material discovery, and AI techniques applied to chemistry and physics.

Toronto has a growing quantum ecosystem. We had an amazing incubator called The Creative Destruction Lab that churns out many quantum technology startups. We also had companies like Xanadu and Zapata Computing that are actively working in quantum computation. There have been many collaborations coming out of meetings between these groups and the University of Toronto.

On Creating The Meta-VQE Algorithm

This work belongs to a big family known as variational quantum algorithms — mixing quantum and classical computation.

Source: https://arxiv.org/pdf/2009.13545.pdf

  • For every quantum state in a quantum computer, we measure some properties of the state.

  • Then, depending on those properties' values and a series of parameters that must be optimized classically, we put these values in a classical optimizer.

  • Then, the optimizer proposes a new series of parameters for the quantum circuit.

  • Next, we close the loop and repeat the loop many times until convergence.

  • Finally, after convergence, we obtained the final ground-state energy profile of a Hamiltonian. An example of a Hamiltonian is a molecule, for instance.

The Meta-VQE learns the full Hamiltonian in the following manner:

  • Instead of running the VQE algorithm for each parameter of the Hamiltonian, we run the algorithm for a particular training point, learn the encoding of these Hamiltonians into the quantum system, and reran the algorithm at test time to obtain the final energy profile.

  • The goal is to save computational time and build a machine learning application for quantum simulation.

On Developing The Tequila Library

Source: https://github.com/aspuru-guzik-group/tequila

Tequila is a unified library implemented at a high level. Users can translate their code into many other possible quantum languages. Tequila uses other hardware devices and simulators as the backend. It is intended to be a general platform that unifies the languages and benchmarks results.

We need collaborators who want to implement their own backends into Tequila. As an open-source language, it has a Github repository for anyone to check and contribute.

On Quantum Calling

The expansion of quantum computing is new. There are many theoretical and experimental challenges to attack. Because of that, we need people from different backgrounds. On the experimental side, we need engineers, physicists, and material designers to build quantum devices and implement quantum gates. On the theoretical side, we need quantum scientists, mathematicians, financial analysts, biologists, and chemists to develop algorithms that solve challenges in their domains.

From the software point of view, we need a good interface between classical and quantum computers. We need quantum languages and quantum simulators to benchmark our experiments before running them on a real computer. Thus, software engineers are more than welcome to this field.

In the end, there are spaces for everyone because we are developing a new tool.

On Encouraging Science

I consider it a part of my job as a scientist is to educate the new generation of scientists. Furthermore, as a woman in STEM, I believe it’s my duty to prove to the new generation that women are welcomed and can do as good of a job as others in the quantum computing field.

I also learned a lot from doing outreach activities. When you have to explain something as complicated as quantum mechanics to someone without any experience (in mathematics and physics even), you need to think carefully and go to the point, which is way harder than I expect.

Timestamps

Her Contact Info

Her Recommended Resources

See this content in the original post

About the show

Datacast features long-form conversations with practitioners and researchers in the data community to walk through their professional journey and unpack the lessons learned along the way. I invite guests coming from a wide range of career paths - from scientists and analysts to founders and investors — to analyze the case for using data in the real world and extract their mental models (“the WHY”) behind their pursuits. Hopefully, these conversations can serve as valuable tools for early-stage data professionals as they navigate their own careers in the exciting data universe.

Datacast is produced and edited by James Le. Get in touch with feedback or guest suggestions by emailing khanhle.1013@gmail.com.

Subscribe by searching for Datacast wherever you get podcasts or click one of the links below:

If you're new, see the podcast homepage for the most recent episodes to listen to, or browse the full guest list.