Analysis and Suppression of Errors in Quantum Random Access Memory under Extended Noise Models

Rohan Mehta, Gideon Lee, Liang Jiang · 2024

Read Analysis and Suppression of Errors in Quantum Random Access Memory under Extended Noise Models by Rohan Mehta, Gideon Lee, Liang Jiang (2024) 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-2412-10318.txt.

Introduction

Rohan Mehta, Gideon Lee, and Liang Jiang extend what is known about error tolerance in quantum random access memory. Their 2024 paper centers on the bucket-brigade design, prized because query errors grow only slowly with memory size under local noise. Past proof covered errors on single routers during a query, a narrow set of cases.

The authors widen the result to settings closer to real hardware: faulty starting states, errors linked across space, and coherent errors. The slow growth of error holds in all cases. Coherent errors raise the bound by up to a square factor, so the authors discuss adapting randomized compiling to soften them. They note reset steps between queries may not be needed.

Catalog record

CREATORRohan Mehta, Gideon Lee, Liang Jiang
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE84 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-10318

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Analysis and Suppression of Errors in Quantum Random Access Memory under Extended Noise Models Rohan Mehta, Gideon Lee, Liang Jiang Abstract Quantum random access memory (QRAM) is required for numerous quantum algorithms and network architectures. Previous work has shown that the ubiquitous bucket-brigade QRAM is highly resilient to arbitrary local incoherent noise channels occurring during the operation of the QRAM [PRX Quantum 2, 020311 (2021)], with query infidelities growing only polylogarithmically with memory width when errors are assumed to only occur on individual routers. We extend this result to a large class of generalized settings that arise in realistic situations, including arbitrary initialization errors, spatially correlated errors, as well as coherent errors, maintaining the polylogarithmic scaling in all instances. Fully quantifying the extent to which QRAM’s noise resilience holds may provide a guide for the design of QRAM architectures — for instance, the resilience to initialization errors indicates that a reset protocol between successive queries may not be necessary. In the case of coherent errors, we find an up-to-quadratic increase in the infidelity bound, and therefore discuss generalizations to randomized compiling schemes, which usually are rendered inapplicable in the QRAM setting, to tailor these errors into more favorable stochastic noise. Preprint: APS/123-QED I Introduction The concerted global effort towards the realization of quantum computers predicates on a quantum advantage that can be realized in comparison to classical devices. Many of the most interesting proposals for quantum algorithms predicate on solving longstanding, fundamentally classical problems: we provide the algorithm a description of a classical problem, and with high probability, we are returned a useful classical answer. Quantum processing can provide utility at intermediary stages to reduce resource consumption, but given the interest in these so-called classical end-to-end applications [ 1 ] , interfaces between classical data and quantum processors are imperative. The primary module to achieve this, quantum random access memory (QRAM), an analogue of classical random access memory, has been presumed as an oracle to facilitate claims of quantum speedup in various algorithms [ 2 , 3 , 4 ] . Moreover, QRAM-based schemes are prevalent in complexity-optimal quantum state preparation [ 5 , 6 ] , quantum data centers [ 7 ] , and resource-efficient metrology [ 8 , 9 ] . Despite the importance of realizing a quantum random access memory, concrete constructions have been explored only recently, and an experimental demonstration of QRAM has yet to be established. Classical RAM retrieves an entry x_{i} from a database \vec{x} of size N=2^{n} corresponding to an n bit binary address state i . Conversely, a QRAM coherently queries a superposition of classical data based on a quantum address. As an oracle, a query to the QRAM acts on a quantum address concatenated with a \ket{+} state bus qubit, \ket{\psi_{\textit{in}}}=\sum_{i=0}^{N-1}\alpha_{i}\ket{i}\ket{+} , where each i\in\mathbb{F}_{2}^{n} is the binary index corresponding to a particular entry in \vec{x} . O:\sum_{i=0}^{N-1}\alpha_{i}\ket{i}^{\mathcal{A}}\ket{+}^{\mathcal{B}}\mapsto\sum_{i=0}^{N-1}\alpha_{i}\ket{i}^{\mathcal{A}}Z^{x_{i}}\ket{+}^{\mathcal{B}} (1) where, for simplicity, each x_{i} is a classical bit of information, hence the need for only one bus qubit. We will define \ket{\psi_{\textit{out}}}:=O\ket{\psi_{\textit{in}}} (2) Here, \mathcal{A},\mathcal{B} denote the address and bus registers, respectively. Though we will not dwell on these applications, modifications to aspects of the prototypical QRAM scheme can be made to accommodate quantum data, which of course cannot be copied but instead transferred due to no-cloning limitations. In order to achieve the claimed advantages of algorithms requiring a QRAM, it is crucial that the QRAM operation can be carried out with high fidelity. Yet, performing full error correction on QRAM is expected to extremely challenging (see for instance [ 10 , 11 , 12 , 13 ] for a review of the outstanding challenge of error-corrected QRAM). Native noise suppression and mitigation is thus critical, as this will reduce the level of error correction necessary for QRAM to reach certain target error rates, ensuring that the overhead for fault-tolerant QRAM does not dominate the cost of scalable quantum computing. Reducing the base error rates of QRAM may also allow us to explore the applications of QRAM in the early fault-tolerance regime, where full error correction may not be available. One prong of this research program is to explore noise resilient experimental designs of the various components that make up a QRAM [ 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 ] . At the same time, from an architectural point of view, it is equally important to identify schemes that ensure that QRAM is highly noise-resilient even prior to error correction [ 22 , 23 , 24 , 25 , 12 ] . To this end, the most promising approach to building QRAM involves the bucket-brigade architecture [ 22 , 23 ] and variations thereof [ 26 , 25 ] . In Section II , we review prior work on the bucket-brigade architecture and its noise resilience, which constitute the primary object of study for this work. For now, it suffices to say that despite requiring a hardware overhead that goes as O(N) , bucket-brigade QRAM only suffers a query infidelity that goes as O(\text{poly}(n)) , and can thus be said to be naturally robust [ 10 ] . Given how much hinges on the bucket-brigade’s natural noise resilience, it is important to fully demarcate the bounds of this noise resilience. We will implicitly assume the bucket-brigade architecture whenever mentioning QRAM. The contributions of this work to the research program of building robust QRAM falls into two main parts. The first part pertains to expanding known results on the natural noise resilience enjoyed by the bucket-brigade QRAM. Extending beyond the standard paradigm of local, uncorrelated, stochastic noise occurring during the operation of the QRAM, in Section III , we greatly extend the scope of this result to show that QRAM also exhibits a natural noise resilience against a much larger class of errors that may arise in practice. This includes: (1) Spatially correlated noise across multiple routers (Section III.2 ), (2) Coherent noise, in which the relative phase of Kraus operators on different routers must be taken into account (Section III.3 ), and (3) Propagated errors between queries that arise from failing to reset the QRAM tree, for instance, due to hardware constraints (Section III.1 ). In all the above cases, we show that QRAM retains its polylogarithmic error scaling, but the strength of the noise suppression depends on the class of the error. Beyond relying on the bucket-brigade’s natural noise resilience, it is also essential to explore alternatives to full QEC that can further push down these error rates. Standard schemes for error mitigation [ 27 ] may be applied [ 28 ] , but do not necessarily serve QRAM well; such schemes primarily produce expectation values , whereas the goal of QRAM is to produce |\psi_{out}\rangle as a resource for computation. The complexity and size of the bucket-brigade architecture is such one must then look to bespoke means of suppressing errors in QRAM [ 24 , 12 , 20 ] . Following this line of research, the second part of our work involves a QRAM-specific proposal for suppressing errors beyond the natural noise resilience that QRAM already enjoys. In Section IV , we consider Pauli twirling [ 29 ] , which

Read the complete text