ArithmerBlog 本資料は2021年4月22日に社内共有資料として展開していたものをWEBページ向けにリニューアルした内容になります。
■Outline
- About D-Wave Machines
- Theory of (Adiabatic) Quantum Annealing
- How to run D-Wave Machines
- Applications of D-Wave Machines
■Reference
■About D-Wave Machines
What is a ”quantum computer” ?
Quantum Computer
- a high-performance computer taking advantage of quantum effects
What are ”quantum annealing” and “D-Wave machine?”
Quantum Annealing
- Quantum algorithm to search the global minimum (最適解) of a combinatorial optimization problem (組み合わせ最適化問題)
From Wikipedia: Quantuma annealing
D-Wave Machine
Computer system implementing quantum annealing on its hardware
- Analog computer
- Sampler (heuristic machine)
- Under a severe environment
From Wikipedia: Quantuma annealing
D-Wave Machines
D-Wave machine is the first commercialized quantum computer.
■Theory of Quantum Annealing
Qubit
D-Wave Machine is composed of qubits (量子ビット) quantum + bit → qubit
Quantum Annealing: Framework
Quantum Annealing: Target Term
Many combinatorial optimization problems can be mapped onto the Ising model
Quantum Annealing: Combinatorial Optimization Problem
Quantum Annealing: Driver Term
Quantum Annealing: GS of Driver Term
Quantum Annealing: Mechanism
Quantum Annealing: Failure of Quantum Annealing
Fast annealing cause state transition from the GS to other states
Quantum Annealing: Difference with Simulated Annealing(Classical)
Quantum annealing can pass through energy barrier
Quantum Annealing: Failure of Simulated Annealing
Fast cooling trap the state in the local minimum
Simulated Annealing:
Quantum Annealing: Theoretical Estimation
Worst Evaluation(最悪評価)
Required time to get optimal solution
Quantum Annealing: Details of Required Time
■How to run D-Wave Machines
D-Wave Cloud Service
D-Wave Leap
- D-Wave Inc. provides D-Wave cloud service called “Leap”
- One need to create a free account
- Leap provides free developer access, free time: one minute (QPU usage at $2000/hour)
- D-Wave provides python SDK called “Ocean” to access to QPU, *QPU: quantum version of CPU
From https://cloud.dwavesys.com/leap/login/?next=/leap/
View of D-Wave Leap (GUI part)
Available D-Wave Machines
Difference between 2000Q and Advantage
Flowchart to run D-Wave Machine
SPIN: -1 or 1
BINARY: 0 or 1
From https://docs.ocean.dwavesys.com/
Preparation to run D-Wave Machines by Python
Coding Process
1. Define model
2. Set up D-Wave Sampler (QPU)
3. Embed and sample Ising or QUBO model
4. Analyze solutions!
Code Example
Minor Embedding
D-Wave machine can not implement a full-connected model directly。
”EmbeddingComposites” automatically do minor embedding
Chain Breaking
Applications of D-Wave Machines: Industry
From dwavejapan.com
Applications of D-Wave machines: Academia
Simulating physics with computersInternational Journal of Theoretical Physics 21, 467– 488(1982)
Observation of topological phenomena in a programmable lattice of 1,800 qubits
https://arxiv.org/abs/1803.02047
Probing the Universality of Topological Defect Formation in a Quantum Annealer:
Kibble-Zurek Mechanism and Beyond
https://arxiv.org/abs/2001.11637 etc…
Characteristic Properties and Future Prospects
Characteristic Properties
- There are hardware problems at each machine.
- Sometimes noise or bias affects results.
Prospects
- There are several researches to speed up.
- Different scheduler may be added. e.g., reverse annealing, pause, quench
- Larger scale, more couplers, more stable
Summary
- D-Wave machines work by quantum annealing algorithm
- Quantum annealing is theoretically faster than classical algorithms, details depend on the type of problem
- There are several generations of D-Wave machines, and what they can do is different.
- We can easily access and use D-Wave machines through Leap
