Reproduce “ResQ: A Novel Framework to Implement Residual Neural Networks on Analog Rydberg Atom Quantum Computers” 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
Nicholas S. DiBrita, Jason Han, Tirthak Patel
RESQ presents a framework for mapping Residual Neural Networks, interpreted as Neural Ordinary Differential Equations, onto the continuous Hamiltonian dynamics of analog Rydberg atom quantum computers. This approach demonstrates a 50% average accuracy improvement over classical neural networks with comparable parameters and exhibits strong resilience to real hardware noise on the QuEra Aquila system for classification tasks.
Key insights
- Analog Rydberg atom quantum computers, with their intrinsic continuous-time Hamiltonian evolution, offer a natural and efficient platform for implementing continuous neural network models like Neural ODEs, unlike discrete gate-based systems.
- Input features can be scaled linearly with the number of qubits by encoding them into local detuning coupling strengths, making the approach scalable for larger datasets.
- The specific parameterization of time-dependent laser pulses enables both data embedding and trainable model parameters within the system's physical evolution, leading to parameter-efficient learning.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.