Slides

QIC710 Presentation_Colin.pdf

A very short abstract

Quantum machine learning (QML) extends the pool of hardware for machine learning (ML) by utilizing the quantum computer. In this presentation, we briefly introduce QML (with an emphasis on supervised learning using neural networks), and explain why it is challenging to benchmark its practical power with the tools we have today. These challenges call for us to examine if proving quantum advantages over classical machine learning should continue to dominate the literature as it does today. Lastly, we provide some examples for alternative research questions.

Outline

1. Intro to Classical Neural Networks & Quantum Neural Networks
   1. QML Pipeline (Encoding, Applying a Model, Measurement, Optimization)
2. Potential problems about framing QML research around “beating” classical machine learning
   1. How should we formalize learning?
   2. Upheavals on learning theory from deep learning
   3. How do we define the quantum advantage?
3. Alternative research agendas
   1. Searching for a quantum perceptron: what are the natural building blocks for QML algorithms?
   2. Computing gradients of quantum circuits: how to make quantum software ready for ML applications?

List of references

• M. Schuld and N. Killoran, "Is quantum advantage the right goal for quantum machine learning?," arXiv preprint arXiv:2203.01340, 2022. (https://arxiv.org/pdf/2203.01340.pdf)
• IBM Qiskit Textbook (https://qiskit.org/textbook-beta/summer-school/quantum-computing-and-quantum-learning-2021)
• Schuld, M., & Petruccione, F. (2021). Machine Learning with Quantum Computers. Springer. (https://link.springer.com/book/10.1007/978-3-030-83098-4)
• Abbas, A., Sutter, D., Zoufal, C. et al. The power of quantum neural networks. Nat Comput Sci 1, 403–409 (2021). https://doi.org/10.1038/s43588-021-00084-1 (https://www.nature.com/articles/s43588-021-00084-1)
• S. Aaronson, "Read the fine print," Nature Physics, vol. 11, no. 4, pp. 291-293, 2015/04/01 2015, doi: 10.1038/nphys3272.