Training deep quantum neural networks

Add to your list(s) Download to your calendar using vCal

• Kerstin Beer (Leibniz Universität Hannover)
• Friday 15 October 2021, 11:00-12:00
• Virtually, at Zoom.

If you have a question about this talk, please contact .

Machine learning, particularly as applied to deep neural networks via the back-propagation algorithm, has brought enormous technological and societal change. With the advent of quantum technology it is a crucial challenge to design quantum neural networks for fully quantum learning tasks. In my talk I will present a truly quantum analogue of classical neurons and explain how to use it to form a quantum feed-forward neural networks capable of universal quantum computation. For training these networks we use the fidelity as a cost function and benchmark the proposal for the quantum task of learning an unknown unitary operation. We find remarkable generalization behavior and robustness to noisy training data. My talk will be based on a recent work of us [1]. For digging deeper in to the topic after the talk I would recommend reading about finding an optimal lower bound on the probability that such a trained network gives an incorrect output for a random input [2] and about considering graph-structured quantum data for training our quantum neural networks [3].

[1] https://www.nature.com/articles/s41467-020-14454-2 [2] https://arxiv.org/abs/2003.14103 [3] https://export.arxiv.org/abs/2103.10837

Where: Virtually on Zoom https://us02web.zoom.us/j/88908652048?pwd=MDV3N3k0YnNWMlhKOEk1NDZlUEtaUT09

This talk is part of the Cavendish Quantum Information Seminar Series series.

This talk is included in these lists:

• Virtually, at Zoom

Note that ex-directory lists are not shown.