Deep Learning for Quantum Computers
Problem being addressed
Deep learning has had a profound impact on the machine learning community over the past few years. But, how does that translate to the scientific forefront of Quantum Computing?
Bolzmann machines are a class of deep learning algorithms that have seen much application over the past few years in various spheres including winning the famous Netflix prize for the best algorithmic recommender. In this research, the ideas of Bolzmann machines are translated from classical computing to quantum computing. Two new quantum algorithms (1) gradient estimation with quantum sampling and (2) quantum amplitude estimation are proposed. Further, it is shown that the problem of training a Bolzmann machine is equivalent to quantum state preparation. In terms of speed-up experimental results show that the quantum approach is quadraticly better than training a classical Bolzmann machine.
Advantages of this solution
Quantum computing lies at the forefront of science and given the recent developments by teams working at Google and IBM it seems that fully fledged quantum computers are not too far away. This research represents a breakthrough that takes deep learning to the quantum arena.
Solution originally applied in these industries
Possible New Application of the Work
Aerospace & Defence Sector
It could be anticipated that quantum deep learning has applications to cryptography and other areas in the defense sector. Quantum computing is expected to have huge ramifications for cryptography which is central to the defence sector.
Deep learning with Bolzmann machines has applications in solving combinatorics problems in the distribution industry. Quantum deep learning approaches (when they can scale) could probably solve these better.
One potential application of quantum computers is in simulating physical systems. Quantum deep learning could be used in this regard to enhance our understanding and engineering at the smallest of scales: the fabrication of quantum devices.
Source DOI: #############
Source URL: #############