Source-linked AI summary

Circuit-Based Quantum Random Access Memory for Classical Data

Daniel K. Park, Francesco Petruccione, June-Koo Kevin Rhee

arXiv:1901.02362v1quant-ph

TL;DR

Quantum algorithms need an efficient way to encode classical data into quantum states, while prepared states cannot generally be reused after measurement. The paper presents circuit-based FF-QRAM for constructing and updating quantum databases, requiring O(n) qubits and O(Mn) steps, and introduces quantum forking to reduce QRAM queries in certain applications.

  • Problem

    Quantum information processing requires efficient conversion of classical data into quantum states, while prepared QRAM inputs cannot generally be reused after measurement or copied.

  • Method

    The paper develops circuit-based flip-flop QRAM to construct and update quantum databases, with post-selection enabling continuous data encoding as probability amplitudes.

  • Results

    O(n) qubits and O(Mn) flip-register-flop steps register n-bit classical data with arbitrary probability amplitudes across M memory cells.

  • Takeaways & Limitations

    Quantum forking can reduce QRAM queries in certain applications while enabling independent unitary processes in superposition.

  • Takeaways & Limitations

    Post-selection adds a data-dependent runtime factor of ∼1/P(1), although preprocessing can improve success probability in some instances.

Abstract

from arXiv · show

A prerequisite for many quantum information processing tasks to truly surpass classical approaches is an efficient procedure to encode classical data in quantum superposition states. In this work, we present a circuit-based flip-flop quantum random access memory to construct a quantum database of classical information in a systematic and flexible way. For registering or updating classical data consisting of $M$ entries, each represented by $n$ bits, the method requires $O(n)$ qubits and $O(Mn)$ steps. With post-selection at an additional cost, our method can also store continuous data as probability amplitudes. As an example, we present a procedure to convert classical training data for a quantum supervised learning algorithm to a quantum state. Further improvements can be achieved by reducing the number of state preparation queries with the introduction of quantum forking.

Introduction

Efficient classical-to-quantum data encoding is needed for quantum algorithms, but practical schemes must also address implementation assumptions and repeated state preparation. The paper proposes circuit-based FF-QRAM and introduces quantum forking to reduce QRAM queries in certain applications.

  • Efficiently generating arbitrary quantum input states is necessary for practical quantum algorithms processing complex classical data.
  • The scheme’s practicality depends on inactive routing elements not introducing noticeable errors.
  • FF-QRAM uses standard circuit-based computation without a routing algorithm to write, update, and superpose unsorted classical data with non-uniform amplitudes.The construction uses O(n) qubits and O(Mn) quantum operations for M entries represented by n bits.
  • Quantum database states generally cannot be reused after measurement or copied, making QRAM preparation costly across algorithms with identical inputs.
  • Quantum forking is introduced to split unitary processes in superposition and reduce QRAM queries in certain applications.

Results

The flip-flop QRAM constructs quantum databases from classical data with circuit-based operations, supports amplitude encoding and updates, and extends state preparation through quantum forking. Its main trade-offs are post-selection overhead and potentially dominant register-operation costs.

  • Quantum database construction: A QDB represents classical bit strings in computational-basis states, with probability amplitudes encoding continuous data or normalized data occurrence.For M n-bit data entries, n + m qubits are sufficient, where m relates to the address register.
  • Flip-flop QRAM: The FF-QRAM uses classically controlled Pauli-X gates and n-qubit controlled rotations to selectively flip, rotate, and restore bus states while registering successive data entries.The process repeats across entries and can assign non-uniform weights; selective flip-flop operations also support updates to existing databases.
  • Amplitude encoding and overhead: Post-selection matches rotation angles to desired amplitudes but increases total runtime by a data-dependent factor of ∼1/P(1).Pre-processing can improve the success probability in some distance-based classification instances; post-selection is unnecessary for some binary-data transformations.
  • Implementation considerations: Register-operation gates can dominate total QRAM runtime, making efficient implementation of the multi-qubit controlled rotation critical for practicality.The paper notes that circuit-depth reductions depend on the gate set available in a specific experimental setup.
  • Application to quantum support vector machine: O(log2(MN)) hardware resources and O(MN) flip-register-flop operations construct the quantum database for the quantum support vector machine.The circuit writes training-data components through rotations determined by elements of the training vector, after post-selecting the register outcome |1⟩R.
  • Quantum forking: Quantum forking applies independent unitary processes in superposition, reducing QRAM queries for certain applications and enabling real-part inner-product estimation.With an arbitrary ancilla, the described method uses O(n) additional gates and reduces QRAM queries by a factor of ∼1/2; adding a phase gate accesses the imaginary part.

Discussion

The paper proposes a circuit-based flip-flop QRAM for encoding classical data flexibly, while identifying post-selection, resource overhead, and state reusability as practical considerations.

  • Architecture: FF-QRAM registers n-bit classical data with arbitrary probability amplitudes using O(n) qubits and O(Mn) flip-register-flop steps.It supports discrete information as qubits and continuous information as probability amplitudes.
  • Architecture: The architecture is implemented with standard quantum circuits, providing compatibility with quantum computing techniques without relying on a routing algorithm.
  • Resource considerations: Register-operation overhead can dominate the runtime beyond the O(Mn) flip-register-flop steps.
  • Quantum forking: Quantum forking addresses QDB non-reusability after measurement and can reduce QRAM queries in some applications, including inner-product evaluation.Finding further applications with even constant query reductions remains open.

Methods

The circuit implementation decomposes multi-controlled operations into Toffoli gates and ancillas, with merged operations reducing depth and an alternative error model reducing the noise-exposed time steps.

  • Multi-controlled gates: CnNOT is recursively decomposed into 2n − 3 Toffoli gates and n − 2 ancilla qubits prepared in |0⟩.Additional Toffoli gates uncompute the ancillas so they become unentangled from the system.
  • Circuit depth: The ¯cX operations between consecutive controlled rotations can be merged, reducing circuit depth under the stated simultaneous-operation assumption.
  • Gate accounting: Writing M classical data entries of length n uses n(M + 1) single-qubit gates, while each CnRy(θ) uses two single-qubit gates and two CnNOT gates.
  • Noise exposure: The CnNOT implementation exposes 2n − 1 qubits to 2⌈log2(n)⌉ − 1 time steps.
  • Noise exposure: If each CnRy(θ) requires only n + 1 independent errors, the noise-exposed time becomes (n + 1)M + n(M + 1).
Loading 1901.02362v1…