Ding Liu1,2, Shi-Ju Ran2,3, Peter Wittek4,5,6,7, Cheng Peng8, Raul Blázquez García2, Gang Su8,9 and Maciej Lewenstein2,10
Published 30 July 2019 • © 2019 The Author(s). Published by IOP Publishing Ltd on behalf of the Institute of Physics and Deutsche Physikalische Gesellschaft
1 School of Computer Science and Technology, Tianjin Polytechnic University, Tianjin 300387, People's Republic of China
2 ICFO-Institut de Ciencies Fotoniques, The Barcelona Institute of Science and Technology, E-08860 Castelldefels (Barcelona), Spain
3 Department of Physics, Capital Normal University, Beijing 100048, People's Republic of China
4 University of Toronto, M5S 3E6 Toronto, Canada
5 Creative Destruction Lab, M5S 3E6 Toronto, Canada
6 Vector Institute for Artificial Intelligence, M5G 1M1 Toronto, Canada
7 Perimeter Institute for Theoretical Physics, N2L 2Y5 Waterloo, Canada
8 School of Physical Sciences, University of Chinese Academy of Sciences, Beijing 100049, People's Republic of China
9 Kavli Institute for Theoretical Sciences, and CAS Center of Excellence in Topological Quantum Computation, University of Chinese Academy of Sciences, Beijing 100190, People's Republic of China
10 ICREA, Passeig Lluis Companys 23, E-08010 Barcelona, Spain
Abstract
The resemblance between the methods used in quantum-many body physics and in machine learning has drawn considerable attention. In particular, tensor networks (TNs) and deep learning architectures bear striking similarities to the extent that TNs can be used for machine learning. Previous results used one-dimensional TNs in image recognition, showing limited scalability and flexibilities. In this work, we train two-dimensional hierarchical TNs to solve image recognition problems, using a training algorithm derived from the multi-scale entanglement renormalization ansatz. This approach introduces mathematical connections among quantum many-body physics, quantum information theory, and machine learning. While keeping the TN unitary in the training phase, TN states are defined, which encode classes of images into quantum many-body states. We study the quantum features of the TN states, including quantum entanglement and fidelity. We find these quantities could be properties that characterize the image classes, as well as the machine learning tasks.