Opportunities and challenges in near-term quantum computers

Transcription

1 Opportunities and challenges in near-term quantum computers Alejandro Perdomo-Ortiz Senior Research Scientist, Quantum AI Lab. at NASA Ames Research Center and at the University Space Research Association, USA Honorary Senior Research Associate, Computer Science Dept., UCL, UK Scientific advisor, Cambridge Quantum Computing Ltd. (CQCL), UK Most recent work: - Perdomo-Ortiz, Feldman, Ozaeta, Isakov, Zhu, O Gorman, Katzgraber, Diedrich, Neven, De Kleer, Lackey, and Biswas. On the readiness of quantum optimization machines for industrial applications, arxiv preprint arxiv: , Perdomo-Ortiz, Benedetti, Realpe-Gomez, and Biswas. Opportunities and challenges for quantumassisted machine learning in near-term quantum computers. arxiv: (2017). To appear in the Quantum Science and Technology (QST) invited special issue on What would you do with a 1000 qubit device? Fujitsu Laboratories Advanced Technology Symposium 2017 Mountain View, CA, October 11, 2017

2 Research approach Quantum-enhanced applications Application focus areas - Combinatorial optimization: Fault Diagnosis, Protein Folding, Bayesian Networks, etc - Machine Learning: unsupervised and semisupervised generative modeling QC programming Physics Insights Simulation tools Analytical methods Outcomes from application investigations - Programming and parameter setting of real hardware device characterization - Hybrid quantum-classical approaches - Future QC architectural design elements - Application-specific and general classical solvers Physical insights into and intuitions for QC

3 Classical pre- and post-processing Quantum processing Inference New hybrid proposal that works directly on a low-dimensional representation of the. Compressed Hidden layers Raw input Training samples Visible units Hidden units Qubits Measurement

4 Classical pre- and post-processing Quantum processing Inference New hybrid proposal that works directly on a low-dimensional representation of the. Quantum sampling Compressed Hidden layers Classical generation or reconstruction of Raw input Training samples Visible units Hidden units Qubits Generated samples Measurement

5 Classical pre- and post-processing Quantum processing Inference New hybrid proposal that works directly on a low-dimensional representation of the. Quantum sampling Compressed Hidden layers Classical generation or reconstruction of Raw input Corrupted image Reconstructed image Visible units Hidden units Qubits Measurement Benedetti, Realpe-Gomez, and Perdomo-Ortiz. Quantum-assisted Helmholtz machines: A quantum-classical deep learning framework for industrial sets in near-term devices. arxiv: (2017).

6 Lesson 1: Focus on the hardest problems of interest to ML experts (e.g., generative models in unsupervised learning). Quickest path to demonstrating quantum advantage in the near-term Lesson 2: Focus on novel hybrid quantum-classical approaches. Cope with hardware constrains arxiv: (2017). To appear in the Quantum Science and Technology (QST) invited special issue on What would you do with a 1000 qubit device?

7 Support slides

8 - Most previous proposed work have highly optimized powerful classical counterparts (e.g., on discriminative/classification tasks) State-of-the-art QML - Need for qram (case of most gatebased proposal). - Qubits represent visible units; issue for case of large sets Lesson 1: Move to intractable problems of interest to ML experts (e.g., generative models in unsupervised learning). Potential impact across social and natural sciences, engineering, and more Bayesian inference Deep learning Probabilistic programming Others... Hypothesis: intractable sampling problems enhanced by quantum sampling

9 Lesson 2: Need for novel hybrid approaches. Challenges solved: Benedetti, et al. Estimation of effective temperatures in quantum annealers for sampling applications: A case study with possible applications in deep learning. PRA 94, (2016). Benedetti, et al. Quantum-assisted learning of graphical models with arbitrary pairwise connectivity. arxiv: (2016). Ex.: Restricted Boltzmann Machines (RBM) RBM Hidden units, u Benedetti, et al. Quantum-assisted Helmholtz machines: A quantum-classical deep learning framework for industrial sets in near-term devices. arxiv: (2017). Computationally bottleneck Where, Visible units, v Widely used in unsupervised learning Perdomo-Ortiz, et al. Opportunities and Challenges in Quantum-Assisted Machine Learning in Near-term Quantum Computer. arxiv: (2017).