Using Frame Randomization to Mitigate Errors in Quantum Optimization
Rachel E. Johnson, Joshua A. Job, Steve Adachi · 2025
Read Using Frame Randomization to Mitigate Errors in Quantum Optimization by Rachel E. Johnson, Joshua A. Job, Steve Adachi (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-14142.txt.
Introduction
Rachel Johnson, Joshua Job, and Steve Adachi tested error mitigation for quantum optimization in twenty twenty-five. Near-term quantum hardware is noisy, and two techniques — universal frame randomization and randomized compilation — insert random twirling gates that average errors down without deepening the circuit. The team applied both to the quantum approximate optimization algorithm on a superconducting device.
Their target was the ground state of a frustrated Ising ring. Randomized circuits from both methods pulled the estimated energy closer to the noiseless value than the unprotected run. Randomized compilation landed nearer the ideal than frame randomization in this test, though with wider spread. The study gives hardware users a concrete comparison of two cheap defenses for a flagship algorithm.
Catalog record
| CREATOR | Rachel E. Johnson, Joshua A. Job, Steve Adachi |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 15 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-14142
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Using Frame Randomization to Mitigate Errors in Quantum Optimization Preprint: APS/1 Rachel E. Johnson, Joshua A. Job, Steve Adachi Abstract Error mitigation is essential for near-term quantum devices, and one promising technique is frame randomization. This method inserts random twirling gates into a circuit to reduce errors while preserving unitarity and depth. We apply frame randomization to the quantum approximate optimization algorithm (QAOA) with p = 1 p=1 on a superconducting quantum circuit system, demonstrating its potential to improve energy calculations. Specifically, we investigate the use of QAOA to calculate the lowest energy state of a frustrated Ising ring system and compare the results of randomized circuits generated using two different randomized techniques. Our results show that both methods can mitigate errors, with expected extremal energy values of 5.25 ± 0.145 5.25\pm 0.145 and 4.08 ± 0.36 4.08\pm 0.36 , for Randomized Compilation and Pauli frame randomization respectively, compared to 2.63 ± 0.068 2.63\pm 0.068 without randomization and 5.676 ± 0.006 5.676\pm 0.006 with a noiseless simulator. I Introduction Quantum computing has the potential to revolutionize various fields, but it is hindered by the presence of noise in quantum systems. To mitigate this noise, techniques such as quantum error correction, suppression, and mitigation are employed qct Nielsen and Chuang 2000 Rieffel and Polak 2011 . Quantum error correction is impractical for current noisy intermediate-scale quantum (NISQ) devices due to high error rates Preskill 2018 . Therefore, in this paper we focus on an error mitigation strategy, frame randomization, which has been shown to reduce errors in quantum computations Cai et al. 2023 Wallman and Emerson 2016 . In this paper, we explore the application of frame randomization to the Quantum Approximate Optimization Algorithm (QAOA) for solving a frustrated Ising ring problem. I.1 Frame Randomization Pauli frame randomization and Randomized Compilation are very similar error mitigation techniques that mitigate errors associated with operating two-qubit gates Wallman and Emerson 2016 . Here we will refer to the general approach that encompasses both techniques as frame randomization . To apply frame randomization to a specific circuit, we start by drawing the circuit diagram and framing different cycles within the circuit. For example, in the circuit in Figure 1 , we define 4 cycles and each cycle in this example contains no more than one gate operation for each qubit. 0 1 2 H H H X T Z H H H Figure 1: Circuit and cycles for randomized compilation on a simple quantum circuit. Each cycle is circled with a dashed line. Once the cycles have been defined, each cycle must be framed by inserting random single-qubit gates on either side of each qubit in the cycle. This is depicted in Figure 2 . These random gates are chosen in such a way that the overall circuit unitary stays the same. For example, if a Hadamard is randomly chosen to frame one side of gate, another Hadamard is inserted on the other side of the gate. When a qubit is measured, we do not insert randomized gates for that cycle. 0 1 2 H H H X T Z H H H Figure 2: Circuit with randomized frames around each cycle. Arbitrary random gates are inserted before and after each cycle to “frame” the cycle. At this step, the circuit has become much deeper, however because of the choice of randomized gates, the circuit can always be compiled to preserve overall circuit depth after compiling Sivarajah et al. 2020 . Two slight variations of the general technique of frame randomization are Pauli frame randomization, which inserts random Pauli gates, and Randomized Compilation, which inserts random Clifford gates. Wallman and Emerson 2016 . We will investigate Pauli Frame Randomization by using the TKET software package and Randomized Compilation by using the True-Q software package. I.2 The Ising Model The Ising model describes a system of connected nodes that can be in one of two states. We can describe the energy of this system with the following Hamiltonian where J is a matrix representing interactions between nodes and h represents an external magnetic field \mathcal{H}=-\sum_{\braket{i,j}}J_{i,j}\sigma_{i}\sigma_{j}-\sum_{i}h\sigma_{i}. (1) In this model, frustration occurs when at least one edge must be violated, meaning the state chosen for the two adjacent nodes do not minimize the energy of that coupling term. Figure 3: Frustrated Ising ring Roberts et al. 2020 . Nodes are connected by coupling terms denoted J , J_{L} , and -J_{R} . The negative coupling term causes frustration. Red nodes are in one state (spin up) and the blue node is in the other state (spin down). This can occur in a ring where all but one of the couplings are positive, but the last coupling is negative as shown in Figure 3 . In the lowest energy state of this system, the nodes must choose states where they do not agree with the sign of at least one of the couplings. Often frustration can lead to a system with multiple distinct ground states. The frustrated Ising model poses significant challenges for quantum annealers, particularly for certain parameter choices Roberts et al. 2020 . This motivated our investigation into the performance of QAOA on this problem. I.3 Quantum Approximate Optimization Algorithm (QAOA) The Quantum Approximate Optimization Algorithm, or QAOA, is a hybrid quantum-classical algorithm that can be used to find approximate solutions to combinatorial optimization problems Farhi et al. 2014 Zhou et al. 2020 . It is a discrete approximation of quantum annealing that can be run on gate-based quantum computers Farhi et al. 2000 . A QAOA circuit consists of p layers. In each layer, we apply a phase operator parameterized by \gamma and then a mixing operator parameterized by \beta Zhou et al. 2020 . An example QAOA circuit is shown in Figure 4 . This circuit prepares a state as shown below where \mathcal{H}_{C} represents the cost Hamiltonian and \mathcal{H}_{M} represents the mixer Hamiltonian. \ket{\Psi(t)}=e^{-i\beta\mathcal{H}_{C}}e^{-i\gamma\mathcal{H}_{M}}\dots e^{-i\beta\mathcal{H}_{C}}e^{-i\gamma\mathcal{H}_{M}}\ket{\Psi(0)} Here, we will set p=1 and evaluate the circuit for a range of \gamma and \beta hyper-parameters. This will allow us to examine errors. We can create a map of the expected energy values obtained from running each pair of \gamma and \beta values. We will refer to this as the energy landscape. 0 1 2 3 e^{-i\gamma_{1}\mathcal{H}_{C}} e^{-i\beta_{1}\mathcal{H}_{M}} e^{-i\gamma_{2}\mathcal{H}_{C}} e^{-i\beta_{2}\mathcal{H}_{M}} \cdots \cdots \cdots \cdots e^{-i\gamma_{p}\mathcal{H}_{C}} e^{-i\beta_{p}\mathcal{H}_{M}} Figure 4: Circuit diagram for QAOA. Gates parameterized by \beta and \gamma are iterated over the length of the circuit. The first complete layer of the circuit is circled with a dashed line. II Experimental Setup We investigate the use of QAOA to solve for the lowest energy state of a 12-node frustrated Ising ring. To simplify the problem and focus on the coupling terms, we can set all h terms to zero. Then the Hamiltonian in Equation 1 is determined only by the coupling terms. To model a ring, we choose adjacent coupling terms (i.e. between nodes (0,1), (1,2), (2,3), etc) to be 1, and all other coupling terms to be zero. Furthermore, to introduce frustration to the ring, we set one of these pairs to have a coupling term of -1. We choose to focus on frustrated rings of 12 nodes because this matches the heavy hexagon layout