Speaker
Description
Imaginary-time evolution (ITE) underpins a broad family of quantum algorithms for quantum simulation, optimization, and ground-state preparation. Yet all existing ITE methods are fundamentally limited by asymptotic convergence, with the energy error decaying only exponentially in imaginary time. We introduce augmented imaginary-time evolution (AITE), a generalized framework that extends conventional ITE by replacing the standard energy-gradient flow with a geometrically informed descent that exploits the higher-order statistical structure of the instantaneous energy distribution. The resulting flow fundamentally changes the convergence dynamics: AITE strictly outperforms standard ITE throughout the entire evolution and exhibits two distinct regimes, consisting of an initial superlinear convergence followed by an extinction regime in which the energy error vanishes exactly at a finite imaginary time. This finite-time convergence contrasts sharply with the asymptotic behavior of conventional ITE and establishes a qualitatively new class of imaginary-time optimization algorithms. Standard ITE is recovered as the zero-skewness limit of AITE, showing that it arises as a special case within the proposed framework. Since ITE underlies a wide range of methods in quantum simulation, variational optimization, quantum machine learning, and computational many-body physics, AITE has broad implications across these areas.
| Topic | Quantum algorithms for machine learning applications |
|---|---|
| Additional Topic | Hybrid quantum-classical approaches for learning and optimization |