Using Reinforcement Learning to find Efficient Qubit Routing Policies for Deployment in Near-term Quantum Computers

Abstract

This paper addresses the problem of qubit routing in first-generation and other near-term quantum computers. In particular, it is asserted that the qubit routing problem can be formulated as a reinforcement learning (RL) problem, and that this is sufficient, in principle, to discover the optimal qubit routing policy for any given quantum computer architecture. In order to achieve this, it is necessary to alter the conventional RL framework to allow combinatorial action space, and this represents a second contribution of this paper, which is expected to find additional application, beyond the qubit routing problem addressed herein. Numerical results are included demonstrating the advantage of the RL-trained qubit routing policy over using a sorting network.

CNOT circuit extraction for topologically-constrained quantum memories

“…However, this approach comes with an enormous overhead in terms of 2-qubit operations, each of which introduces a great deal more noise than a single qubit operation on most realistic architectures [5]. More sophisticated approaches incorporate techniques from computer aided design [8] and machine-learning [9] in order to minimise the extra operations needed by making good choices of initial and intermediate memory locations for the qubits involved. Nevertheless, these are simply refinements of the basic 'search and swap' approach.…”

Nuwa: A Quantum Circuit Transpiler Based on a Finite-Horizon Heuristic for Placement and Routing

“…This optimization problem is often modelled as a placement and routing problem of finding a mapping of the abstract qubits to the physical qubits (placement), followed by iterations of performing entangling gates between qubits that are far apart by moving them closer to each other (routing), for example, via Swap gates. The placement and routing problem has been extensively studied in the context of NISQ algorithms [3,4,7,[10][11][12][13][14][15][16][17][18][19][20][21][22][23], as well as in the fault-tolerant compilation of quantum algorithms using concatenated error-correcting codes [24], lattice surgery-based fault-tolerant quantum computing (FTQC) [7,25], and defect-based FTQC [26].…”

“…Some of the solutions relevant to our approach are those based on the observation that the placement and routing problem is an inherently temporal problem over the span of multiple decision epochs. This includes an exact, but exponentially expensive, dynamic programming solution introduced in [13], temporal planning and constraint pro-gramming [16,19], and methods that use reinforcement learning [14,23]. On one hand, these approaches are superior to the greedy techniques that neglect the future impact of decisions made at earlier decision epochs.…”

Timing and resource-aware mapping of quantum circuits to superconducting processors

“…Different solutions have been proposed to map quantum circuits onto NISQ processors. [20][21][22][23][24][25][26][27][28] map quantum algorithms onto processors with a 2D grid structure. [29][30][31][32][33][34][35] and [36,37] propose mapping algorithms targeting IBM and Rigetti processors, respectively.…”