High-Quality Thermal Gibbs Sampling with Quantum Annealing Hardware

Nelson, Jon; Vuffray, Marc; Lokhov, Andrey Y.; Albash, Tameem; Coffrin, Carleton

doi:10.1103/physrevapplied.17.044046

Cited by 14 publications

(9 citation statements)

References 56 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Explicitly, for the spin glass with a configuration space of size , we use 10,000 reads, which lead to around 300 (for ) to 5,000 (for ) distinct configurations in the subset S (the actual numbers fluctuate due to the non-deterministic behavior). We note that the concept of the rescaling of coupling constants has been previously used to effectively change the temperature of the distribution generated by the quantum annealer 18 , 19 , which is however not required here.…”

Section: Resultsmentioning

confidence: 99%

“…The results show that irrespective of the choice of the rescaling parameter a , the low temperature magnetization always coincides with the theoretical expectation, which is obtained from a brute force sampling of the partition function. Hence, a is here not used as a method to tune the effective temperature, as compared to the approaches mentioned above 18 , 19 . As discussed above, a smaller value of a leads to sampling of more excited states, and consequently the better the agreement with the theoretical prediction also for higher temperatures.…”

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Efficient low temperature Monte Carlo sampling using quantum annealing

Sandt

Spatschek

2023

Sci Rep

View full text Add to dashboard Cite

Quantum annealing is an efficient technology to determine ground state configurations of discrete binary optimization problems, described through Ising Hamiltonians. Here we show that—at very low computational cost—finite temperature properties can be calculated. The approach is most efficient at low temperatures, where conventional approaches like Metropolis Monte Carlo sampling suffer from high rejection rates and therefore large statistical noise. To demonstrate the general approach, we apply it to spin glasses and Ising chains.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Resultsmentioning

confidence: 99%

Efficient low temperature Monte Carlo sampling using quantum annealing

Sandt

Spatschek

2023

Sci Rep

View full text Add to dashboard Cite

show abstract

“…Today's largest and most mature quantum annealers are produced by D-Wave Systems with a qubit technology based on superconducting loops [15][16][17][18][19]. There is still an ongoing research effort to probe the effectiveness of the D-Wave hardware as an optimization tool [20][21][22][23], as well as a Gibbs sampler [24][25][26][27][28][29][30][31][32][33], which could prove useful in machine learning applications. With the latest systems from D-Wave featuring thousands of qubits, it remains important to identify the noise sources and their effects on the output statistics in order to discover further use cases for the hardware.…”

Section: Introductionmentioning

confidence: 99%

“…To model fluctuations in the programmed Hamiltonian, we use randomness in its parameters. Mixtures of random Hamiltonians with longitudinal field noise has been shown to play a major role in explaining the formation of the anomalous single qubit response to magnetic field changes [31,32] and of the effective spurious links [31,33].…”

Section: Introductionmentioning

confidence: 99%

Signatures of Open and Noisy Quantum Systems in Single-Qubit Quantum Annealing

Morrell

Vuffray

Lokhov

et al. 2022

Self Cite

View full text Add to dashboard Cite

We propose a quantum annealing protocol that more effectively probes the dynamics of a single qubit on D-Wave's quantum annealing hardware. This protocol uses D-Wave's h-gain schedule functionality, which allows the rapid quenching of the longitudinal magnetic field at arbitrary points during the anneal. This features enables us to distinguish between open and closed system dynamics as well as the presence or absence of longitudinal magnetic field noise. We show that both thermal and magnetic field fluctuations are key sources of noise that need to be included in an open quantum system model to reproduce the output statistics of the hardware.

show abstract

“…That is, it has an additional output that indicates if a generated samples was created successfully or if it has to be discarded. Indeed, quantum algorithms for learning and inference of specific probabilistic models have been proposed, including quantum Bayesian networks [22], quantum Boltzmann machines [2,19,37,42], and Markov random fields [40,4,25]. However, many of these methods are either approximate or require so-called fault-tolerant quantum computers-a concept that cannot yet be realized with the state-of-the-art quantum hardware.…”

Section: Introductionmentioning

confidence: 99%

On Quantum Circuits for Discrete Graphical Models

Piatkowski¹,

Zoufal²

2022

Preprint

View full text Add to dashboard Cite

Graphical models are useful tools for describing structured high-dimensional probability distributions. Development of efficient algorithms for generating unbiased and independent samples from graphical models remains an active research topic. Sampling from graphical models that describe the statistics of discrete variables is a particularly challenging problem, which is intractable in the presence of high dimensions. In this work, we provide the first method that allows one to provably generate unbiased and independent samples from general discrete factor models with a quantum circuit. Our method is compatible with multi-body interactions and its success probability does not depend on the number of variables. To this end, we identify a novel embedding of the graphical model into unitary operators and provide rigorous guarantees on the resulting quantum state. Moreover, we prove a unitary Hammersley-Clifford theorem-showing that our quantum embedding factorizes over the cliques of the underlying conditional independence structure. Importantly, the quantum embedding allows for maximum likelihood learning as well as maximum a posteriori state approximation via state-of-the-art hybrid quantum-classical methods. Finally, the proposed quantum method can be implemented on current quantum processors. Experiments with quantum simulation as well as actual quantum hardware show that our method can carry out sampling and parameter learning on quantum computers.Preprint. Under review.

show abstract

High-Quality Thermal Gibbs Sampling with Quantum Annealing Hardware

Cited by 14 publications

References 56 publications

Efficient low temperature Monte Carlo sampling using quantum annealing

Efficient low temperature Monte Carlo sampling using quantum annealing

Signatures of Open and Noisy Quantum Systems in Single-Qubit Quantum Annealing

On Quantum Circuits for Discrete Graphical Models

Contact Info

Product

Resources

About