Toward Large-Scale Distributed Quantum Long Short-Term Memory with Modular Quantum Computers

TL;DR

This work tackles the scalability challenge of quantum long short-term memory by introducing a distributed QLSTM that uses modular quantum computing to partition variational quantum circuits across multiple QPUs. By embedding VQCs inside LSTM gates and partitioning these circuits into smaller subcircuits, the approach enables large-scale temporal modeling on near-term devices while maintaining a hybrid quantum-classical training loop via the parameter-shift rule. Empirical results on a damped harmonic oscillator and NARMA sequences demonstrate stable training dynamics and accurate time-series predictions, with competitive performance compared to centric QLSTM configurations across varying qubit resources. The findings highlight the potential of modular, distributed quantum architectures to scale quantum sequence modeling and pave the way for integration into quantum HPC ecosystems, albeit with ongoing work needed in partitioning strategies and noise mitigation.

Abstract

In this work, we introduce a Distributed Quantum Long Short-Term Memory (QLSTM) framework that leverages modular quantum computing to address scalability challenges on Noisy Intermediate-Scale Quantum (NISQ) devices. By embedding variational quantum circuits into LSTM cells, the QLSTM captures long-range temporal dependencies, while a distributed architecture partitions the underlying Variational Quantum Circuits (VQCs) into smaller, manageable subcircuits that can be executed on a network of quantum processing units. We assess the proposed framework using nontrivial benchmark problems such as damped harmonic oscillators and Nonlinear Autoregressive Moving Average sequences. Our results demonstrate that the distributed QLSTM achieves stable convergence and improved training dynamics compared to classical approaches. This work underscores the potential of modular, distributed quantum computing architectures for large-scale sequence modelling, providing a foundation for the future integration of hybrid quantum-classical solutions into advanced Quantum High-performance computing (HPC) ecosystems.

Figures (6)

• Figure 1: Split Variational Quantum Circuits (VQCs) functioning as Quantum Neural Networks (QNN), as used in this paper.
• Figure 2: Distributed QLSTM Framework via Split-QNN Cells.
• Figure 3: Training and testing loss curves comparing centric QLSTM models with different qubit counts (3, 4, 5, 6) and a distributed QLSTM approach, illustrating stable convergence for both methods over 100 epochs.
• Figure 4: Time-series comparison of predicted angular velocities $\dot{\theta}(t)$ from the centric QLSTM (3, 4, 5, 6 qubits) and the distributed QLSTM relative to ground truth, demonstrating each model’s ability to capture both oscillatory and damped behavior.
• Figure 5: Training and testing loss curves for the centric QLSTM (3, 4, 5, 6 qubits) and the distributed QLSTM, demonstrating rapid convergence to near-zero loss for both approaches.