QHack2021: Hackathon Recap and project development experience
On the 17th of February, the Xanadu team hosted its second quantum machine learning hackathon, QHack2021. Unlike the first hackathon which was held at Xanadu’s HQ in Toronto, this year, the event took place virtually, from February 17 till 26, giving the opportunity to experts and enthusiasts in quantum computing and quantum machine learning to participate from all around the world.
In this article, I’m going through my experience in the hackathon, starting from finding a team, to tackling the QML Challenges, and submitting the final project at the end.
Before the event
One week before the hackathon kicked off, I managed to create a team. I used both the Xanadu Slack and the Unitary Fund Discord servers to reach out to other participants, within a few hours several participants responded, and the next day our team was formed, the “Entangled_Nets” team.
Although I’ve joined a couple of hackathons before, QHack was the first time I had an international collaboration. All three of us joined from different countries, and although it was extremely exciting to have people with different perspectives, the time zone difference was challenging. We had three different time zone with a 6+ hours difference, but since we created the team 6 days before the event starts, we had enough time to figure out a way to manage this issue. Besides, in these 6 days, we held multiple meetings in which we discussed the ways we should prepare for the QML challenges and approach the final project.
Quantum machine learning - PennyLane
We're entering an exciting time in quantum physics and quantum computation: near-term quantum devices are rapidly…
We started off to be reading tutorials from PennyLane, where each one of us chose a topic that found interesting and read about it. Furthermore, we tried to construct a primary project idea as a preparation for the final project.
The QML Challenges
On the 17th, both the QML Challenges and QHack streaming sessions kicked off, with exciting talks and presentations from experts in the fields of quantum computing and quantum machine learning. There were two sessions a day with about 4 talks each, this lasted for 3 days with a total of 35 hours of live-streamed video with over 13,500 views. The recordings were made available on the QHack Twitch channel for later review.
The QML challenges were prepared to improve the participants' quantum machine learning skills and compete against other teams for the top spot. It contained 12 individual problems, ranging in difficulty from building basic quantum circuits to implementing QML workflows. Here is the list of the challenges presented:
simple_circuits_20, simple_circuits_30, simple_circuits_50
quantum_gradients_100, quantum_gradients_200, quantum_gradients_500
circuit_training_100, circuit_training_200, circuit_training_500
vqe_100, vqe_200, vqe_500
Detailed explanations and templates of the problem sets can be found on the Xanadu QHack GitHub repo.
The solutions to the QML challenges might end up on the PennyLane website, besides there are other tutorials that will help you learn more about QML, so make sure to check it out.
After 2 days, we’ve submitted all the problems thanks to my team members Edo and Andrei(Voicu Tomut)’s great efforts, and we ranked 15th on the scoreboard. Thus we granted our way to the Open Hackathon. Moreover, we were awarded the $250 AWS credit by Amazon Braket, in which we were able to access QPU’s of Rigetti, IonQ, and D-Wave with high-performance simulators on the cloud.
The Open Hackathon
Although we had finished the QML challenges, we had 3 days till the Open Hackathon starts, which was a good pause to take a breath and plan for the final project, as we prepare before, we tried to divide the tasks that each one of us is responsible for, and on the 22nd the Open Hackathon kicked off with an introductory presentation to the AWS services.
While working on the project, we submitted an initial entry describing the project and the resource estimation in order to qualify for an additional power-up prize of $4000 AWS credits, which we received on the 24th and was very valuable for us to test the codes on actual quantum hardware.
Although I’ve heard about Amazon braket before, It was my first time using it during the hackathon, In fact, I took a glance at AWS when Xanadu first announced Amazon Braket, by creating an account and checking out how it worked. Right after I created a notebook instance on the server, there were introductory notebooks explaining the platform and how to use it, which was a good exercise to learn more about Amazon Braket and the services it’s providing. For more information on this, check out the Amazon Braket Getting Started page.
Besides, the top 50 teams in the QML Challenge received an API key for the alpha of Sandbox at Alphabet’s Floq API, where teams were able to develop and test their QML creations on Floq’s TPU-based high-performance quantum simulator for the duration of the Open Hackathon portion of the event.
Without diving into the details, our project is titled “Event Classification with Layerwise Learning for Data Re-uploading Classifier in High Energy Physics”. We used the PennyLane— Python library for differentiable programming of quantum computers to train a quantum computer the same way as a neural network.
The project details can be found on the GitHub Repo.
Join the community and stay tuned
QML is an extremely exciting field and the community is constantly expanding. Many people from different backgrounds are involved, and you can always find someone who will help and provide you with the necessary resources.
I would like to thank my teammates, Edo and Andrei(Voicu Tomut), for sharing their knowledge, coming up with unique solutions, and being open to new ideas. I couldn't have come this far without their help.
And Many Thanks to the Xanadu team for organizing an amazing event, and making it possible for everyone to participate and become part of this community.