Xanadu Software Competition: the results are in!
At Xanadu, we have a big focus on open-source software which makes quantum computers more accessible. Our software portfolio has two crown jewels:
- Strawberry Fields: Program photonic quantum computers — an easy-to-use interface for creating photonic quantum circuits, with built-in compilation methods, simulators, and applications. The software interface to Xanadu’s quantum photonic hardware (coming soon!).
- PennyLane: The TensorFlow of quantum computing — a cross-platform library dedicated to machine learning on quantum computers. Runs quantum machine learning algorithms on Rigetti’s Forest, IBM’s Qiskit, Xanadu’s Strawberry Fields, or Microsoft’s QDK, and connects them seamlessly with PyTorch and TensorFlow.
We are constantly updating these libraries with new releases, adding hardware backends, and implementing exciting features. We’ll have more announcements to make in the coming months!
The main purpose of this post is to highlight the winners of Xanadu’s first-ever software competition, demonstrating creative and exciting uses of our software stack. Over the past year, we received a number of submissions across three categories: Research, Education, and Software. Without further ado, we are happy to announce the following award winners and honourable mentions in each category:
Research award
1st place:
Samuel Yen-Chi Chen, Chao-Han Huck Yang, Jun Qi, Pin-Yu Chen, Xiaoli Ma, Hsi-Sheng Goan
Variational Quantum Circuits for Deep Reinforcement Learning
[Paper] [Code]
2nd place:
Michel Barbeau & Joaquin Garcia-Alfaro
Faking and Discriminating the Navigation Data of a Micro Aerial Vehicle Using Quantum Generative Adversarial Networks
[Paper] [Code]
Honourable mention:
Miller Eaton
Gottesman-Kitaev-Preskill State Preparation by Photon Catalysis
[Paper]
Education award
1st place:
Dawid Kopczyk
Variational Quantum Circuits
[Notebook]
2nd place:
Jack Ceroni
Continuous QAOA Optimization with Photonic Quantum Computation: A Tutorial
[Blog post]
Software award
1st place:
Roeland Wiersema
Rocky Raccoon: PennyLane and the Quantum Log-Likelihood
[Repository]
2nd place:
Filippo Miatto & Alessandro Luongo
CJTricks: Choi-Jamiolkowski tricks in Strawberry Fields
[Respository]
Honourable mentions:
Marcus Edwards
Juan Leni
William Pol & the scikit-quantum team
First-place winners receive $1000 and second-place winners $500. Congratulations to the top finishers, and thank you to everyone who participated.
We saw some really interesting submissions to the competition (even a poem!). We will highlight some of our judges’ favourites below, with descriptions provided by the winners themselves.
Rocky Racoon by Roeland Wiersema
We reached out to Roeland to summarize his winning entry:
“The goal of Rocky Raccoon is to offer users an easy way to learn a quantum machine learning model by minimizing a very specific cost function; the quantum log-likelihood (QLLH). The QLLH describes the “distance” between the states of a quantum system (described by density matrices). If we make these states dependent on input data and adjustable parameters, we can learn a categorical classifier by minimizing the QLLH. For some data sets, this can lead to improved performance over the classical equivalent of this loss function.
Our choice of controllable quantum system is of course the quantum computer. Rocky Raccoon relies on PennyLane for simulating and optimizing a parametrized quantum system on a quantum computer. By combining this with Tensorflow, a Python deep learning framework, it is easy to train complex hybrid architectures.
I had a lot of fun working on this project, and I really enjoyed the interaction with the PennyLane Software Team on Github and the discussion forum.”
Roeland’s submission also came with an awesome title page! You can see part of it in the banner image at the top of this blog.
Variational Quantum Circuits for Deep Reinforcement Learning by Samuel Yen-Chi Chen and collaborators.
Samuel had the following to say about the work:
“In this work, we demonstrated that variational quantum circuits can be more than classifiers; they can play the role of an action-selector in a real-world scenario. For example, the proposed quantum-based reinforcement learning agent can succeed in dynamic channel selection in a multi-channel environment. Although current quantum devices are still at a small scale, it is expected that larger machines with more qubits will be released soon and we hope to scale up this work and tackle more sophisticated problems.
Thanks to Xanadu for offering this competition and the software platform, empowering many researchers and benefiting the community. And thanks for the generous help from Prof. Hsi-Sheng Goan in NTU Physics Dept and NTU-IBM Q Hub. I especially thank Dr. Pin-Yu Chen from IBM Research and also thanks to my colleagues Chao-Han Huck Yang, Jun Qi, and Prof. Xiaoli Ma at Georgia Institute of Technology, School of ECE; they shared their experience in deep reinforcement learning and helped to define and set up a real-world case for dynamic telecommunication networks.”
CJ Tricks by Filippo Miatto and Alessandro Luongo
Here’s a description of the work, in the winners’ own words:
“In our research, we use Strawberry Fields to develop new quantum devices. When we started using it, it was clear that it was a tool unlike anything we had ever used before. This encouraged us to hack it and test its limits, and we noticed that we could feed any matrix as initial state (as opposed to just normalised state vectors).
That’s when we had the idea to feed the identity matrix: the result would be the channel itself, according to the Choi-Jamiolkowsky isomorphism. With a few tweaks and a lot of index gymnastics we contributed to the code with a utility to extract unitary and non-unitary representations of quantum channels, just by feeding identities!
This is a classical example of how a ‘bug’ is a powerful feature in disguise.”
Thanks again to everyone who submitted to the competition!
For those who missed out on the competition deadline, but still want to explore quantum software, Xanadu is hosting QHACK’19, a quantum machine learning hackathon, in November.
Further details can be found at qhack.ai
