Contents/conteúdo

Mathematics Department Técnico Técnico

Quantum Computation and Information Seminar  RSS

Sessions

10/05/2016, 14:00 — 15:00 — Abreu Faro Amphitheatre
, Google Inc.

Latest Developments at the Google Quantum Artificial Intelligence Lab

I will describe two architectures, quantum annealers and quantum circuits, that the Quantum AI team at Google is currently developing in order to accelerate tasks important for AI. Quantum annealers are a promising tool to find good solutions to hard combinatorial optimization problems. In recent benchmarking we were able to demonstrate that finite range quantum tunneling enables the D-Wave 2X quantum annealer to solve crafted benchmark problems $10^8$ times faster than thermal annealing that does not employ tunneling. We are now studying whether speedups are also available for generic problems such as Boolean satisfiability problems and find this is indeed the case. I will discuss the implications of these studies for the design of next generation quantum annealers. The second class of quantum processors we are developing are quantum circuits. Those were initially devised as an architecture to achieve digital error corrected quantum computation. In the near term however when the number of physical qubits is still small quantum circuits have to be operated as an analog device. Yet, with high fidelity gate operations and low measurement error it is possible to achieve quantum supremacy over all known classical algorithms with just about 50 qubits. The talk will conclude with an outlook how to apply quantum resources to enhance artificial intelligence. As an example of how to use quantum annealing in machine learning, I will describe learning from very noisy data. Using the quantum circuits we implemented what could be described as a quantum neural network. In a first application, we used such a circuit to calculate the energy surface of molecular hydrogen to chemical precision.

Joint session with the Physics of Information Colloquium. Please note exceptional day, time and room.

Supported by: Phys-Info (IT), SQIG (IT), CeFEMA and CAMGSD, with funding from FCT, FEDER and EU FP7, specifically through the Doctoral Programme in the Physics and Mathematics of Information (DP-PMI), FCT strategic projects PEst-OE/EEI/LA0008/2013 and UID/EEA/50008/2013, IT project QuSim, project CRUP-CPU CQVibes, the FP7 Coordination Action QUTE-EUROPE (600788), and the FP7 projects Landauer (GA 318287) and PAPETS (323901).

 

Instituto de TelecomunicaçõesCAMGSDFCT7th Framework Programme