Zizhu Wang - Solving quantum problems with variational optimization with deep generative networks

Solving quantum problems with variational optimization with deep generative networks

This seminar, given by Zizhu Wang, will happend on 20 February 2026, at 13:0. It will take place in Room 24-25/405.

Find a map of the campus here.

Abstract

The Variational Generative Optimization Network (VGON) provides a model-agnostic variational optimization framework that utilizes deep generative models to solve difficult quantum optimization problems. Unlike methods that directly tune parameters, VGON learns a mapping of probability transport from a latent space to the solution space. This enables the parallel and efficient generation of a large number of high-quality, highly diverse approximate optimal solutions on classical hardware, such as GPUs.  We apply VGON to four quantum tasks: (1) Identifying optimal high-dimensional quantum states for entanglement detection; (2) Finding the ground states of one-dimensional many-body spin models via variational quantum circuits, while avoiding the barren plateau problem; (3) Generating an orthogonal basis for the degenerate ground-state space of many-body Hamiltonians—specifically, diverse degenerate ground states—in a single training run; and (4) Generating robust pulse sequences for superconducting quantum gates.