Speakers
Description
Quantum Signal Processing (QSP) is a powerful approach for generating and approximating univariate polynomials but is limited by increasing circuit depth and parity restrictions. We introduce Weighted Quantum Signal Processing (WQSP), a generalization of QSP that incorporates fixed weights into signal operators, with standard QSP recovered as the unit-weight case. Classical weight encoding enables efficient realization of high-degree polynomials with linear to exponentially reduced circuit resources. We apply this approach to realize activation functions in Kolmogorov–Arnold Networks (KANs) for multivariate function approximation. WQSP-based KANs outperform Fourier KANs, QSP-KANs, and other KAN variants while providing a compact, and resource-efficient architecture for polynomial and function realization suitable for NISQ-compatible neural architectures.
| Topic | Quantum algorithms for machine learning applications |
|---|---|
| Additional Topic | Quantum variational circuits and their applications |