Bulletin of the American Physical Society
APS March Meeting 2023
Volume 68, Number 3
Las Vegas, Nevada (March 5-10)
Virtual (March 20-22); Time Zone: Pacific Time
Session Z70: Algorithms for Quantum Annealers and Analog ComputersFocus
|
Hide Abstracts |
Sponsoring Units: DQI Chair: Davide Venturelli, NASA QuAIL - USRA Room: Room 409 |
Friday, March 10, 2023 11:30AM - 11:42AM |
Z70.00001: Privacy-preserving machine learning with tensor networks Alejandro Pozas-Kerstjens Vast amounts of data are routinely processed in machine learning pipelines, every time covering more aspects of our interactions with the world. When the models are made public, is the safety of the data used for training them guaranteed? This is a crucial question, especially when processing sensitive data such as medical records. The state-of-the-art protection techniques, despite being deployed commercially, consist in adding noise at some stage during the training process, and thus imply a tradeoff between privacy protection and performance. |
Friday, March 10, 2023 11:42AM - 11:54AM |
Z70.00002: Simulating a low-weight encoding of the Fermi-Hubbard Model with QAOA Joseph Barreto, Ryan Levy, Lucas Brady, Zoe Gonzalez Izquierdo, Jeffrey S Marshall, Eleanor G Rieffel, Zhihui Wang, Filip Wudarski Recent work [1] has proposed a compact fermion-to-qubit mapping for the Fermi-Hubbard model (FHM), potentially lowering the overhead needed to simulate such a system on quantum hardware. In prior work [2], we numerically investigated how well a quantum annealer could prepare ground states of the FHM encoded using the resource-efficient mapping. In order to compare these results to gate-model systems, we turn to a QAOA-style approach and replace the annealing schedule with both Trotterizations thereof and also more general Hamiltonian-variational ansatze based on the terms of the mapped FH Hamiltonian. We investigate the achievable ground state fidelities of these gate sequences, consider various optimizations of the ansatze structure, and offer resource estimations of the optimized QAOA protocols for near-term hardware. |
Friday, March 10, 2023 11:54AM - 12:06PM |
Z70.00003: IMPROVING QUANTUM STATE PREPARATION TECHNIQUES USING EIGENVECTOR CONTINUATION Anjali Agrawal, Akhil Francis, Alexander F Kemper Quantum state preparation is one of the key steps in many quantum algorithms, and numerous algorithms for state preparation have been proposed and studied in order to reduce the cost and improve efficiency. A number of important limitations remain, such as poor scalability, barren plateaus in parameter optimization, and difficulties with circuit implementation. Here, we consider eigenvector continuation (a quantum subspace expansion method) as a potential solution to these issues. Eigenvector continuation uses ground and/or excited states from a training set in parameter space to form a subspace in which a reduced Hamiltonian can be diagonalized.As eigenvector continuation relies on the preparation of ground states at different points in parameter space, it could be susceptible to the same issues that plague other state preparation algorithms. However, we show that eigenvector continuation is resistant to these kinds of issues, enabling the use of fewer iterations and shallower circuits. We demonstrate this within the context of several spin models namely XY model, XXZ interactive model and XXZ frustrated triangular lattice model by using eigenvector continuation based on states prepared via state of the art techniques: Quantum Imaginary Time Evolution (QITE), adiabatic time evolution, and adapt-VQE. In all cases, we find that an approximate state preparation is sufficient to span the subspace, leading to efficient implementations on NISQ hardware. |
Friday, March 10, 2023 12:06PM - 12:18PM |
Z70.00004: Quantum-inspired classical algorithm for molecular vibronic spectra Changhun Oh, Youngrong Lim, Yat Wong, Bill Fefferman, Liang Jiang We have recently seen the first plausible claims for quantum advantage using Gaussian boson sampling. The obvious next step is to channel the potential quantum advantage to solve practical problems rather than conducting proof-of-principle experiments. An interesting proposal of using a boson sampler is to generate molecular vibronic spectra, which is essentially an estimation of the sum of output probabilities of a boson sampler and is an important problem in chemistry. Also, it has been thought to be difficult for a classical computer to efficiently solve. In this talk, we provide a quantum-inspired classical algorithm to solve the problem as accurately as running a boson sampling. In particular, the classical algorithm enables us to achieve the same accuracy efficiently for the case of Fock-state boson sampling and Gaussian boson sampling. Finally, we provide a different type of molecular vibronic spectra problem for which we might be able to take advantage of boson sampling to solve a classically hard problem, which is also chemically well-motivated. The details can be found in Ref. [1] |
Friday, March 10, 2023 12:18PM - 12:30PM |
Z70.00005: Programmable adiabatic demagnetization for systems with trivial and topological excitations Anne Matthies, Achim Rosch, Mark Rudner, Erez Berg Preparing the ground state of a many-body Hamiltonian on a quantum device is of central importance, both for quantum simulations of molecules and materials, and for a variety of quantum information task. We propose a simple, robust protocol to prepare a low-energy state of an arbitrary Hamiltonian on a quantum computer. The protocol is inspired by the “adiabatic demagnetization” technique, used to cool solid state systems to extremely low temperatures. The adiabatic cooling protocol is demonstrated via an application to the transverse field Ising model. We use half of the qubits to model the system and the other half as a bath. Each bath spin is coupled to a system spin. In a strong magnetic field, the bath spins are prepared in the polarized ground state. By an adiabatic downward sweep of the magnetic field, we change the energy of the bath spins and allow for resonant processes that transfer entropy from the system to the bath qubits. After each cycle, the bath is reset to the ground state. |
Friday, March 10, 2023 12:30PM - 12:42PM |
Z70.00006: A density-matrix renormalization group algorithm for simulating quantum circuits with a finite fidelity Thomas Ayral In this talk, I will introduce a recently developed density-matrix renormalization group (DMRG) algorithm for the simulation of quantum circuits (https://arxiv.org/abs/2207.05612). This algorithm can be seen as the extension of time-dependent DMRG from the usual situation of hermitian Hamiltonian matrices to quantum circuits defined by unitary matrices. Like an actual quantum computer, the quality of the DMRG results is characterized by a finite fidelity. However, unlike a quantum computer, the fidelity depends strongly on the quantum circuit considered. For the most difficult possible circuit for this technique, the so-called "quantum supremacy" benchmark of Google Inc., we find that the DMRG algorithm can generate bitstrings of the same quality as the seminal Google experiment on a single computing core. For a more structured circuit used for combinatorial optimization (Quantum Approximate Optimization Algorithm or QAOA), we find a drastic improvement of the DMRG results with error rates dropping by a factor of 100 compared with random quantum circuits. Our results suggest that the current bottleneck of quantum computers is their fidelities rather than the number of qubits. |
Friday, March 10, 2023 12:42PM - 12:54PM |
Z70.00007: The Landscape of Max-Cut QAOA Lie Algebras Sujay S Kazi, Martin Larocca, Robert Zeier, Marco Farinati, Marco Cerezo, Patrick J Coles It is conjectured that the existence of barren plateaus in the cost function for deep variational quantum algorithms is closely tied to the dimension of the dynamical Lie algebra (DLA) obtained from the set of generators of the ansatz. To this end, given a simple graph G = (V, E), we study the DLAs of different circuit ansatzes aimed at generating the ground state of $H_p$, the Hamiltonian whose ground states are solutions to the Max-Cut problem on G, including but not limited to circuit ansatzes based on automorphism orbits and subsets of common degree, which naturally incorporate symmetries coming from parity and automorphisms of G. We investigate a variety of salient features of these DLAs, such as their linear symmetries, quadratic symmetries, decomposition into simple Lie algebras, and irreducible representation structure. We find that there are many cases of symmetries that cannot be explained by parity and automorphisms alone, and there are additionally many cases of failure of subspace controllability. However, we also present numerical evidence that these effects, while quite complex, become proportionally less frequent as the graph size increases, even for fairly small graphs, making it very likely that the vast majority of graphs have Max-Cut DLA dimensions that scale exponentially with the number of vertices. |
Friday, March 10, 2023 12:54PM - 1:06PM |
Z70.00008: Exploring magnetic phases in a qubit Penrose quasicrystal Alejandro Lopez-Bezanilla The fundamental dynamics of magnetic quasicrystals under an external magnetic field remains a challenging problem that could potentially impact the field design of information processing devices. Using a quantum annealer we built a qubit magnetic quasicrystal based on a Penrose tiling to inspect multiple magnetic textures at the individual-spin level driven by particular selections of physical parameters. In our experiments a broad range of frustrated magnetic configurations are observed as a result of the dynamical activation of some spins while others remain static. Static spin structure factors reveal ferromagnetic and ferrimagnetic modulations compatible with a wide range of spin textures. This study demonstrate that inducing multiple spins arrangements in one single object would offer the prospect of applications in magnetic memory storage by encoding multiple magnetic states in a same geometry. |
Friday, March 10, 2023 1:06PM - 1:18PM |
Z70.00009: Solving Continuous Non-convex Optimization Problems Using a Coherent Optical Network Farhad Khosravi, Ugur Yildiz, Artur Scherer, Pooya Ronagh Real-world optimization problems frequently involve decision variables that are naturally continuous. Common discrete classical optimization techniques and various realizations of coherent Ising machines (CIM) normally require the binarization of the continuous variables. These reformulations typically cause significant errors, are very costly in terms of computational resources, and can result in energy landscapes that are difficult to optimize. We propose modifications to the CIM dynamics that allow the optical pulses in the ring cavity to behave like soft spins, providing a means for solving continuous non-convex optimization problems natively in a coherent optical network. We demonstrate the capabilities of the proposed technology by presenting a time-to-solution scaling analysis for solving the NP-hard box-constrained quadratic programming (BoxQP) problem. With the BoxQP problem serving as a case study, we propose a coherent continuous-variable machine (CCVM) that we benchmark against state-of-the-art heuristic solvers to solve this problem. We show that by pumping the optical parametric oscillator pulses of a standard CIM less aggressively than usual when solving binary-variable optimization problems, their mean-field amplitudes are able to converge to fractional values. |
Friday, March 10, 2023 1:18PM - 1:30PM |
Z70.00010: Semiclassical Approximate Optimization Algorithm Peter K Schuhmacher, Aditi Misra-Spieldenner, Tim Bode, Tobias Stollenwerk, Dmitry Bagrets, Frank K Wilhelm In this work, we develop a new Semiclassical Approximate Optimization Algorithm (SAOA) as a classical counterpart to the Quantum Approximate Optimization Algorithm (QAOA). This algorithm substitutes a quantum evolution of QAOA by classical spin dynamics. Within the Trotterization scheme of QAOA, this dynamics can be found exactly for any number of layers p defining the algorithm. We test SAOA for the Sherrington-Kirkpatrick (SK) model and the number partition problem - it delivers an approximated optimum with accuracy of order 1/√N or higer in a polynomial time and outperforms QAOA in these cases. |
Friday, March 10, 2023 1:30PM - 1:42PM |
Z70.00011: QAOA algorithms creates pseudo-Boltzmann pure states Juan Jose Garcia-Ripoll, Diego Porras, Pablo Díez-Valle In this talk I will discuss the nature of the pure states created by single-layer and multi-layer Quantum Approximate Optimization Algorithm (QAOA) on universal Ising spin models. I will show that already a single layer produces thermal-like states with Gaussian perturbations [1]. We find that these pseudo-Boltzmann states can not be efficiently simulated on classical computers according to state-of-art techniques, and we relate this distribution to the optimization potential of QAOA. Moreover, we observe that the temperature depends on a hidden universal correlation between the energy of a state and the covariance of other energy levels and the Hamming distances of the state to those energies. As outlook, this work may help develop a better understanding of multi-layer QAOA and adiabatic quantum computation. To be precise, our study has revealed that a single-step QAOA with very small angles approximates well a thermal distribution. Given that single-layer QAOA approximates a short-time step in adiabatic quantum evolution, it is natural to expect that one may relate a full adiabatic quantum protocol to a process that creates pseudo-Boltzmann states—with temperatures that now also will depend on the speed of the passage. |
Friday, March 10, 2023 1:42PM - 2:18PM |
Z70.00012: Leveraging quantum walks and spin frustration for computation and error mitigation Invited Speaker: Viv Kendon I will give an overview of recent work on using quantum walks for solving optimisation problems, and related work on error mitigation. Quantum walks can solve the same problems as adiabatic quantum computing, but the mechanisms are very different, with excited states populated, and many repeats of fast dynamics replacing slow time evolution. For problems such as MAX2SAT and spin glass ground states, the control parameters do not need to be set precisely to obtain an advantage over naive classical algorithms. Heuristic estimates of control parameters are sufficient, making practical applications viable for similar real world problems. However, many open questions remain around the conditions that generate fast dynamics in quantum walks, and the best classical algorithms are still competitive with quantum, especially SAT solvers. Hybrid strategies offer the best route to exploiting quantum solvers, combining different quantum mechanisms with classical solvers for maximum efficiency overall. For larger problems, the fixed precision of hardware control limits how accurately the problem can be represented. Using copies connected with anti-ferromagnatic links can increase the effective precision by counteracting errors and inhibiting error propagation. Hardware graphs with triangles that can natively display frustration are thus a desirable feature for future hardware design. |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700