Reproduce “Quantum Long Short-term Memory with Differentiable Architecture Search” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Samuel Yen-Chi Chen, Prayag Tiwari
Researchers from Wells Fargo and Halmstad University developed DiffQAS-QLSTM, an automated framework that integrates differentiable quantum architecture search into Quantum Long Short-Term Memory networks. This method enables simultaneous, gradient-based optimization of both quantum circuit parameters and architecture, consistently outperforming manually designed QLSTM baselines in various time-series prediction tasks by achieving lower Mean Squared Error.
Key insights
- Differentiable Quantum Architecture Search provides an efficient and scalable approach for designing quantum-enhanced recurrent neural networks, overcoming limitations of discrete QAS methods.
- Adaptive parameter learning within the Variational Quantum Circuits is crucial for robust performance, as models with fixed parameters showed significantly degraded results.
- Allowing each candidate quantum circuit within the search space to have its own distinct, trainable parameters (NonShared configuration) yields optimal predictive performance for QLSTM models.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.