Quantum Effect or HPC without FLOPS Lugano March 23, 2016
Electronics April 19, 1965 2016 D-Wave Systems Inc. All Rights Reserved 2
Moore s Law 2016 D-Wave Systems Inc. All Rights Reserved 3
www.economist.com/technology-quarterly/2016-03-12/aftermoores-law 2016 D-Wave Systems Inc. All Rights Reserved 4
Predictions for the End of Moore s Law 2016 D-Wave Systems Inc. All Rights Reserved 5
Moore s Law s Law The number of people predicting the death of Moore s law doubles every two years. Peter Lee, a vice-president at Microsoft Research 2016 D-Wave Systems Inc. All Rights Reserved 6
Richard Feynman 1960 1970 1980 1990 2000 2010 2020 2016 D-Wave Systems Inc. All Rights Reserved 7
What is a Quantum Computer? Exploits quantum mechanical effects Built with qubits rather than bits Operates in an extreme environment Enables quantum algorithms to solve very hard problems Quantum Processor 2016 D-Wave Systems Inc. All Rights Reserved 8
Characteristics of Classical Digital Systems Binary Separable Barriers 2016 D-Wave Systems Inc. All Rights Reserved 9
Quantum Effects Superposition Entanglement Quantum Tunneling 2016 D-Wave Systems Inc. All Rights Reserved 10
Quantum Turing Machine 1960 1970 1980 1990 2000 2010 2020 2016 D-Wave Systems Inc. All Rights Reserved 11
Algorithms David Deutsch (1992): Determine whether f: {0,1} n {0,1} is constant or balanced using a quantum computer Daniel Simon (1994): Special case of the abelian hidden subgroup problem Peter Shor (1994): Given an integer N, find its prime factors Lov Grover (1996): Search an unsorted database with N entries in O(N 1/2 ) time 1960 1970 1980 1990 2000 2010 2020 2016 D-Wave Systems Inc. All Rights Reserved 12
Quantum Information Science Quantum key distribution Quantum information processing One-way/ cluster state Topological Quantum Sensor Emerging Quantum Cryptography Quantum Adiabatic Computing Quantum Communication Gate Model Emerging 2016 D-Wave Systems Inc. All Rights Reserved 13
Linear Optics Quantum Computation (LOQC) Bristol University UK LOQC is a paradigm of universal quantum computation using photons as information carriers, mainly linear optical elements including beam splitters, phase shifters, and mirrors to process quantum information, and uses photon detectors and quantum memories to detect and store quantum information. 2016 D-Wave Systems Inc. All Rights Reserved 14
Measurement Based Quantum Computer (MBQC) The one-way or measurement based quantum computer (MBQC) is a method of quantum computing that first prepares an entangled resource state, usually a cluster state or graph state, then performs single qubit measurements on it. It is "one-way" because the resource state is destroyed by the measurements. Experimental Realization of One-Way Quantum Computing with Two-Photon Four-Qubit Cluster States University Heidelberg The outcome of each individual measurement is random, but they are related in such a way that the computation always succeeds. In general the choices of basis for later measurements need to depend on the results of earlier measurements, and hence the measurements cannot all be performed at the same time. 2016 D-Wave Systems Inc. All Rights Reserved 15
Topological Quantum Computer A topological quantum computer is a theoretical quantum computer that employs twodimensional quasiparticles called anyons, whose world lines cross over one another to form braids in a three-dimensional spacetime (i.e., one temporal plus two spatial dimensions). These braids form the logic gates that make up the computer. Alexei Kitaev proposed topological quantum computation in 1997. Experiments in fractional quantum Hall systems indicate these elements may be created in the real world using semiconductors made of gallium arsenide at a temperature of near absolute zero and subjected to strong magnetic fields. 2016 D-Wave Systems Inc. All Rights Reserved 16
MIT group proposes Adiabatic QC 1960 1970 1980 1990 2000 2010 2020 2016 D-Wave Systems Inc. All Rights Reserved 17
Quantum Enhanced Optimization Quantum Hamiltonian is an operator on Hilbert space: H t = E t a i σ i z + b ij σ i z σ j z + Δ t σ i x i i<j i Corresponding classical optimization problem: Obj(a i, b ij ; q i ) = a i q i + b ij q i q j i i<j 2016 D-Wave Systems Inc. All Rights Reserved 18
Energy Landscape Space of solutions defines an energy landscape & best solution is lowest valley Classical algorithms must walk over this landscape Quantum annealing uses quantum effects to go through the mountains 2016 D-Wave Systems Inc. All Rights Reserved 19
Company Background Founded in 1999 World s first quantum computing company Public customers: Lockheed Martin/USC Google/NASA Ames Los Alamos National Lab Other customer projects done via cloud access to systems in D-Wave s facilities 120+ U.S. patents 2016 D-Wave Systems Inc. All Rights Reserved 20
Mission To help solve the most challenging problems in the multiverse: Optimization Machine Learning Monte Carlo/Sampling 2016 D-Wave Systems Inc. All Rights Reserved 21
But, It Is Fundamentally Different Than Anything You ve Ever Done Before! Intel 64 D-Wave Performance (GFLOPS) ~20 (12 cores) 0 Precision (bits) 64 4-5 MIPS ~12,000 (12 cores) 0.01 Instructions 245+ (A-M) 251+ (N-Z) 1 Operating Temp. 67.9 C -273 C Power Cons. 100 w +/- ~0 Devices 4B+ transistors 1000 qubits Maturity 1945-2016 ~1950 s 2016 D-Wave Systems Inc. All Rights Reserved 22
The D-Wave 2X 1000+ qubits Performance: up to 600X Synthetic cases 100,000,000X Power: <25 kw Three orders: Google/NASA LANL Lockheed Martin/USC
D-Wave Container - SCIF-like - No RF Interference 2016 D-Wave Systems Inc. All Rights Reserved 24
System Shielding ~16 layers of shielding between QPU and outside world 2016 D-Wave Systems Inc. All Rights Reserved 25
Processor Environment Cooled to 0.015 Kelvin, 175x colder than interstellar space Shielded to 50,000 less than Earth s magnetic field In a high vacuum: pressure is 10 billion times lower than atmospheric pressure On low vibration floor <25 kw total power consumption for the next few generations 15mK 2016 D-Wave Systems Inc. All Rights Reserved 26
D-Wave 2X Quantum Processor Qubits within red boxes 2016 D-Wave Systems Inc. All Rights Reserved 27
Processing Using D-Wave A lattice of superconducting loops (qubits) Chilled near absolute zero to quiet noise User maps a problem into search for lowest point in a vast landscape which corresponds to the best possible outcome Processor considers all possibilities simultaneously to satisfy the network of relationships with the lowest energy The final state of the qubits yields the answer 2016 D-Wave Systems Inc. All Rights Reserved 28
Programming Model QUBIT q i Quantum bit which participates in annealing cycle and settles into one of two possible final states: 0,1 COUPLER q i q j Physical device that allows one qubit to influence another qubit WEIGHT a i influences the qubit s tendency to collapse into its two possible Real-valued constant associated with each qubit, which final states; controlled by the programmer STRENGTH b ij controls the influence exerted by one qubit on another; Real-valued constant associated with each coupler, which controlled by the programmer OBJECTIVE Obj Real-valued function which is minimized during the annealing cycle Obj(a i, b ij ; q i ) = a i q i + b ij q i q j i ij The system samples from the q i that minimize the objective 2016 D-Wave Systems Inc. All Rights Reserved 29
Programming Environment Operates in a hybrid mode with a HPC System or Data Analytic Engine acting as a co-processor or accelerator D-Wave system is front-ended on a network by a standard server (Host) User formulates problem as a series of Quantum Machine Instructions (QMIs) Host sends QMI to quantum processor (QP) QP samples from the distribution of bit-strings defined by the QMI Results are returned to the Host and back to the user 2016 D-Wave Systems Inc. All Rights Reserved 30
D-Wave Software Environment 2016 D-Wave Systems Inc. All Rights Reserved 31
Discrete Combinatorial Optimization Benchmarks Median Time to Find Best Solution Median time to best solution (s) 10000 Timing Benchmark Smaller is Better CPLEX 1000 METSTABU AKMAXSAT 100 10 1 0.1 0.01 VESUVIUS D-WAVE II 11000 x 0.001 0 100 200 300 400 500 Problem size (number of qubits) 2016 D-Wave Systems Inc. All Rights Reserved 32
Machine Learning: Binary Classification Traditional algorithm recognized car about 84% of the time Google/D-Wave Qboost algorithm implemented to recognize a car (cars have big shadows!) Quantum Classifier was more accurate (94%) and more efficient Ported quantum classifier back to traditional computer, more accurate and fewer CPU cycles (less power)! 2016 D-Wave Systems Inc. All Rights Reserved 33
Google Blog December 8, 2015 When can Quantum Annealing win? Tuesday, December 08, 2015 Posted by Hartmut Neven, Director of Engineering - During the last two years, the Google Quantum AI team has made progress in understanding the physics governing quantum annealers. We recently applied these new insights to construct proof-of-principle optimization problems and programmed these into the D-Wave 2X quantum annealer that Google operates jointly with NASA. The problems were designed to demonstrate that quantum annealing can offer runtime advantages for hard optimization problems characterized by rugged energy landscapes We found that for problem instances involving nearly 1000 binary variables, quantum annealing significantly outperforms its classical counterpart, simulated annealing. It is more than 10 8 times faster than simulated annealing running on a single core. http://googleresearch.blogspot.ca/2015/12/when-can-quantum-annealing-win.html 2016 D-Wave Systems Inc. All Rights Reserved 34
The Most Advanced Quantum Computer in the World 10,000 Number of Qubits 1,000 100 10 16 qubit D-Wave One 128 qubit D-Wave Two 512 qubit D-Wave 2X 1000+ qubit 4 qubit 28 qubit 1 2004 2008 2012 2016 2016 D-Wave Systems Inc. All Rights Reserved 35