Gated QKAN-FWP: Scalable Quantum-inspired Sequence Learning
Researchers have developed a new quantum-inspired sequence learning framework called gated QKAN-FWP, which integrates Fast Weight Programmers (FWPs) with Quantum-inspired Kolmogorov-Arnold Networks (QKANs). This approach utilizes single-qubit data re-uploading circuits as nonlinear activations and incorporates a scalar-gated update rule for stable parameter evolution. The framework demonstrates strong performance in time-series forecasting, including solar cycle prediction, outperforming classical models with significantly more parameters. AI
IMPACT This quantum-inspired approach offers a parameter-efficient alternative for sequence modeling, potentially impacting fields requiring long-term forecasting.