We consider recent works on the simulation of quantum circuits using the formalism of matrix product states and the formalism of contracting tensor networks. We provide simplified direct proofs of many of these results, extending an explicit class of efficiently simulable circuits (log depth circuits with 2-qubit gates of limited range) to the following: let C be any poly sized quantum circuit (generally of poly depth too) on n qubits comprising 1- and 2- qubit gates and 1-qubit measurements (with 2-qubit gates acting on arbitrary pairs of qubit lines). For each qubit line j let D_j be the number of 2-qubit gates that touch or cross the line j i.e. the number of 2-qubit gates that are applied to qubits i,k with i \leq j \leq k. Let D=max_j D_j. Then the quantum process can be classically simulated in time n poly(2^D). Thus if D=O(log n) then C may be efficiently classically simulated.
The study of quantum circuit simulation using classical computers is a key research topic that helps define the boundary of verifiable quantum advantage, solve quantum many-body problems, and inform development of quantum hardware and software. Tensor networks have become forefront mathematical tools for these tasks. Here we introduce a method to approximately simulate quantum circuits using sparsely-populated tensors. We describe a sparse tensor data structure that can represent quantum states with no underlying symmetry, and outline algorithms to efficiently contract and truncate these tensors. We show that the data structure and contraction algorithm are efficient, leading to expected runtime scalings versus qubit number and circuit depth. Our results motivate future research in optimization of sparse tensor networks for quantum simulation.
In this note we describe a simple and intriguing observation: the quantum Fourier transform (QFT) over
Zq, which is considered the most ``quantum'' part of Shor's algorithm, can in fact be simulated efficiently by classical computers. More precisely, we observe that the QFT can be performed by a circuit of poly-logarithmic path-width, if the circuit is allowed to apply not only unitary gates but also general linear gates. Recalling the results of Markov and Shi [MaSh] and Jozsa [Jo] which provided classical simulations of such circuits in time exponential in the tree-width, this implies the result stated in the title. Classical simulations of the FFT are of course meaningless when applied to classical input strings on which their result is already known; Our observation might be interesting only in the context in which the QFT is used as a subroutine and applied to more interesting superpositions. We discuss the reasons why this idea seems to fail to provide an efficient classical simulation of the entire factoring algorithm. In the course of proving our observation, we provide two alternative proofs of the results of [MaSh,Jo] which we use. One proof is very similar in spirit to that of [MaSh] but is more visual, and is based on a graph parameter which we call the ``bubble width'', tightly related to the path- and tree-width. The other proof is based on connections to the Jones polynomial; It is very short, if one is willing to rely on several known results.
We provide numerical evidence that the quantum Fourier transform can be efficiently represented in a matrix product operator with a size growing relatively slowly with the number of qubits. Additionally, we numerically show that the tensors in the operator converge to a common tensor as the number of qubits in the transform increases. Together these results imply that the application of the quantum Fourier transform to a matrix product state with n qubits of maximum Schmidt rank χ can be simulated in O(n (log(n))2 χ 2 ) time. We perform such simulations and quantify the error involved in representing the transform as a matrix product operator and simulating the quantum Fourier transform of periodic states.
We give new bounds on the circuit complexity of the quantum Fourier transform (QFT). We give an upper bound of O(log n+log log(1//spl epsiv/)) on the circuit depth for computing an approximation of the QFT with respect to the modulus 2/sup n/ with error bounded by /spl epsiv/. Thus, even for exponentially small error, our circuits have depth O(log n). The best previous depth bound was O(n), even for approximations with constant error. Moreover, our circuits have size O(n log(n//spl epsiv/)). As an application of this depth bound, we show that P. Shor's (1997) factoring algorithm may be based on quantum circuits with depth only O(log n) and polynomial size, in combination with classical polynomial-time pre- and postprocessing. Next, we prove an /spl Omega/(log n) lower bound on the depth complexity of approximations of the QFT with constant error. This implies that the above upper bound is asymptotically tight (for a reasonable range of values of /spl epsiv/). We also give an upper bound of O(n(log n)/sup 2/ log log n) on the circuit size of the exact QFT modulo 2/sup n/, for which the best previous bound was O(n/sup 2/). Finally, based on our circuits for the QFT with power-of-2 moduli, we show that the QFT with respect to an arbitrary modulus m can be approximated with accuracy /spl epsiv/ with circuits of depth O((log log m)(log log 1//spl epsiv/)) and size polynomial in log m+log(1//spl epsiv/).
We study the classical simulability of a polynomial-size quantum circuit
Cn on
n qubits followed by sparse classical post-processing (SCP) on
m bits, where
m≤n≤poly(m). The SCP is described by a non-zero Boolean function
fm that is classically computable in polynomial time and is sparse, i.e., has a peaked Fourier spectrum. First, we provide a necessary and sufficient condition on
Cn such that, for any SCP
fm,
Cn followed by
fm is classically simulable. This characterization extends the result of Van den Nest and implies that various quantum circuits followed by SCP are classically simulable. Examples include IQP circuits, Clifford Magic circuits, and the quantum part of Simon's algorithm, even though these circuits alone are hard to simulate classically. Then, we consider the case where
Cn has constant depth
d. While it is unlikely that, for any SCP
fm,
Cn followed by
fm is classically simulable, we show that it is simulable by a polynomial-time probabilistic algorithm with access to commuting quantum circuits on
n+1 qubits. Each such circuit consists of at most deg(
fm) commuting gates and each commuting gate acts on at most
2d+1 qubits, where deg(
fm) is the Fourier degree of
fm. This provides a better understanding of the hardness of simulating constant-depth quantum circuits followed by SCP.