Qiskitters Built Amazing Things At This Two-Month Quantum Hackathon
By Ryan F. Mandelbaum, Global Editorial Lead, IBM Quantum & Qiskit
Hackathons might elicit an image of participants working late into the night, surrounded by pizza boxes and empty cola cans while whipping together quick and ingenious projects. But what if a hackathon lasted two months? Participants in the Qiskit Hackathon Europe demonstrated that the extra time could turn quick projects into important research results.
The Qiskit Hackathon Europe: Research Study Groups brought researchers and developers together from April 20 to June 14 to collaborate and build quantum computing projects using Qiskit. Not only did they produce some outstanding work, but they helped further strengthen the Qiskit community and the link between the worlds of physics researchers and software developers.
“The Hackathon was a great opportunity to motivate more advanced members of the quantum computing community to come together,” said Caroline de Groot, PhD student at the Max-Planck Institute of Quantum Optics who co-organized the event. “This was evident through the high quality of submissions, which benefited from the longer time frame. Especially impressive was the promise of these projects for future development and impact on the field, from applications in quantum physics to educational projects.”
Participants with very different backgrounds began by forming teams and pitching ideas in a Discord server, then judges picked the top 20 teams and moved them into the next phase of the event. All participants were able to attend seminars, while the top 20 teams then developed their projects in GitHub repositories, wrote reports, and created video presentations to summarize their work. The top three teams as determined by the judges earned a prize of 3000 €.
“It was very exciting to see many great young minds propose excellent ideas, transform them into projects, and work as very well-oiled teams to produce a final report,” said Fabio Scafrimuto, Europe Team Lead for Community, Education and Outreach in Quantum Computing on the Qiskit team. “I believe that the combination of researchers and developers as well as the long format made this Hackathon a unique experience for both the participants and the organizers.”
So, what can a group of researchers produce with a hackathon’s mentality and two months’ time? Here are the results of the top three winning teams.
Hardware-Efficient VQE
Luciano Pereira, Francisco Escudero, David Fernández, Gabriel Jaumà, and Guillermo F. Peñas teamed up to develop a “library for Qiskit to implement a Hardware-efficient Variational Quantum Eigensolver (VQE) that uses entangled measurements between connected qubits to estimate the energy.” VQE is among the most important near-term quantum algorithms; it employs an iterative process where a classical optimizer updates the parameters of a wavefunction implemented on a quantum computer to find the ground state energy of a Hamiltonian. However, implementing VQE comes with challenges: the finite coherence times and different sources of noise in near-term quantum processors limit the depth of the quantum circuits that developers can implement. Plus, obtaining the ground state energy can require many measurements.
The team set out to find solutions to these challenges using “Hardware-efficient entangled measurements,” or HEEM, measurements that only incorporate entanglement operators between neighboring qubits. Their work both increases the efficiency from a hardware perspective and decreases the number of required measurements. The team reproduced a previous IBM result and received the same solution — but required only a third of the total number of measurements. “We think that our project can expand the range of problems that can be solved with VQE in current devices,” said Luciano Pereira, team member from the Instituto de Física Fundamental in Spain.
Though two months may seam like a long time, the team tried their best not to waste it given the difficulty of their project. They divided the work into smaller pieces which they tackled as sub-groups, then met weekly to share their results and explain their work. However, the long timeline still allowed them to dive even deeper into their problem and search for more creative solutions.
Qiskit provided an intuitive way to implement the hardware-efficient VQE, and Qiskit’s VQE routine served as the basis of the group’s work. On top of this, the team designed an algorithm to construct the HEEM circuits using Qiskit, and then run their code on real quantum hardware.
Now that the hackathon is over, the team is planning to ensure that their code is compatible with other Qiskit functions, such as error mitigation techniques. They also plan on improving their HEEM-constructing algorithm to further reduce the number of measurements required when implementing VQE. They hope that their further work will allow other members of the Qiskit community to understand and employ the algorithm.
Quantum Anomaly Detection
Korbinian Kottmann, Friederike Metz, Joana Fraxanet, and Niccolò Baldelli teamed up to “envision a full-stack self-driving laboratory on a quantum device for the exploration of quantum materials.” Quantum computing researchers hope that one day, quantum processors will be useful for simulating quantum many-body systems — but the field still requires efficient methods to perform these simulations and capture their full complexity. This team created an unsupervised quantum machine learning algorithm that analyzes quantum data in a quantum simulation, and extracts the phase diagram of the system. A phase diagram is essentially a map of the collective behavior of the system based on the environment, sort of like the solid-liquid-gas phase diagram you learn about in chemistry.
The team had to rely on some functions in the Qiskit library in order to implement their algorithm, called Variational Quantum Anomaly Detection (VQAD). VQAD consists of two parts. First comes the Variational Quantum Eigensolver, a function already implemented in Qiskit which the team used and optimized. Second is the anomaly syndrome circuit which the team build from scratch. This circuit is trained to return one value (typically zero) when it receives normal data, and a much larger value when it receives anomalous data. The team also relied on Qiskit’s ability to run noisy simulations on the different models and to experiment with a real device, employing error mitigation techniques and taking the device’s architecture into account.
The event let the team members forge new collaborations within their own research groups and with other institutions. This was especially useful given that the VQAD algorithm combines quantum computing, machine learning, and many-body physics, said team member Joana Fraxanet, Ph.D student from The Institute of Photonic Sciences in Barcelona, Spain. “The combination of people with different backgrounds resulted in a very enriching experience,” she said. “Moreover, I had no experience using Qiskit, and I was interested in this event to experiment a bit with the library, since I believe this is going to be a useful skill for future projects.”
As for the timeline, Fraxanet said the event was intense, but rewarding. “Before starting to work on the project, we tried to set different goals, since it was not clear how ambitious we could be for a two-month project,” she said. “I believe that, in our case, having a very clear idea about what we wanted to do and being organized, setting up regular meetings and deadlines for specific achievements was the key to achieve meaningful and nice results.”
The team believes that the simulation of many-body systems is one of the most promising applications of quantum computing in the near-term, and hopes that their algorithm can inspire similar proposals. They have since put efforts into improving the implementation of their algorithm on real devices, and even wrote a paper containing their results, which you can read on the arXiv here.
Quantum Singular Value Transformation
Bartu Bisgin, Erfan Abedi, Jiri Guth Jarkovsky, Martin Mauser, and Nagme Oruz explored Quantum Singular Value Transformation (QSVT), a “framework that allows one to re-cast complex circuits as the iteration of a ‘universal’ circuit, which also allows for many non-unitary transformations.” QSVT essentially is a meta-algorithm — a way to re-envision and unify many famous algorithms as being based on a more fundamental mathematical structure. Despite being a useful lens through which to study quantum algorithms, the team noticed that the QSVT framework hadn’t received much attention outside of academic study — and they wanted to change that by implementing it in Qiskit.
Bisgin only started studying quantum computing nine months ago as a personal interest, but didn’t feel like other hackathons were accessible, given that one-day events require more experience with software development tools. This was exactly why he and his team members were excited to join the Qiskit Hackathon Europe — it was the opportunity to choose an idea, work on it as a research project, and make a real impact on the quantum computing community.
The team first devised a method to implement the QSVT framework, then worked on how make it accessible to the wider researcher community. The framework is already a challenging one to understand — but thankfully, Qiskit was perhaps the easiest part of the project. “Without Qiskit’s rather meticulous and clear documentation we could have been lost trying to implement the algorithm in the real-world,” Bisgin said.
During the hackathon, the team was able to implement a quantum search algorithm based on the QSVT framework. This is the simplest QSVT-implementable algorithm, according to the team’s project summary, and therefore a good introduction to the framework more generally. The team also ensured that their documentation provided users a deep look into the framework. “We hope that our project serves as an introduction to the beautiful ideas of Quantum Singular Value Transformation for the wider community and that, with it gaining more traction, a lot more people from different backgrounds will be able to explore the topic and potentially come up with groundbreaking stuff,” Bisgin said. “QSVT really is both extremely interesting and promising for gate-based quantum computation.”
The team hasn’t taken their project further, yet, but they’re in touch almost daily and hope to soon implement other QSVT-based algorithms into their library. Bisgin encourages anyone to reach out to him on Qiskit Slack if they’d like to help push the project further.
Congratulations again to all of the winners — and we hope that the rest of the Qiskit community will use and improve upon these contributions to better their own research.
For more stories like these, follow the Qiskit Medium!
