Quantum Long Short-term Memory with Differentiable Architecture Search

TL;DR

This paper introduces DiffQAS-QLSTM, a differentiable framework for optimizing quantum recurrent neural networks, which improves performance in sequence learning tasks by jointly tuning circuit parameters and architecture.

Contribution

It presents the first end-to-end differentiable architecture search method for quantum LSTM models, enabling scalable and adaptive quantum sequence learning.

Findings

Outperforms handcrafted quantum LSTM baselines

Achieves lower loss across diverse tasks

Demonstrates scalability and adaptability in quantum sequence learning

Abstract

Recent advances in quantum computing and machine learning have given rise to quantum machine learning (QML), with growing interest in learning from sequential data. Quantum recurrent models like QLSTM are promising for time-series prediction, NLP, and reinforcement learning. However, designing effective variational quantum circuits (VQCs) remains challenging and often task-specific. To address this, we propose DiffQAS-QLSTM, an end-to-end differentiable framework that optimizes both VQC parameters and architecture selection during training. Our results show that DiffQAS-QLSTM consistently outperforms handcrafted baselines, achieving lower loss across diverse test settings. This approach opens the door to scalable and adaptive quantum sequence learning.