Speaker
Description
We present the first analytical and numerical characterization of double descent in parameterized quantum circuits (PQCs) for regression tasks. By exploiting the Fourier structure of PQC models and the lazy training approximation, we establish a precise connection between overparameterized gradient descent on variational parameters, the quantum neural tangent kernel, and the emergence of double descent. Our results show that the effective model complexity in PQCs is determined by both the number of Fourier modes and the number of variational parameters, leading to a richer landscape of interpolation regimes than in classical linear models. Crucially, the position of the second descent depends on correlations encoded in the circuit Jacobian.
| Topic | Quantum variational circuits and their applications |
|---|---|
| Additional Topic | Theoretical foundations of quantum learning |