Description
The simulation of open quantum systems presents a significant computational challenge due to the exponential scaling of the Hilbert space when solving the Lindblad Master Equation. To circumvent this dimensionality curse, we apply a framework that maps the system dynamics onto a phase-space representation, which we then solve using a variational Ansatz parameterized by an invertible neural network. By integrating this generative architecture with the time-dependent Variational Monte Carlo (t-VMC) method, we demonstrate a scalable approach capable of capturing complex physical phenomena in coupled driven-dissipative bosonic systems across diverse parameter regimes. Our results indicate that this neural-network-based representation maintains high numerical accuracy even in strongly quantum regimes where semi-classical approximations, such as the truncated Wigner method, typically break down. Furthermore, we show that our method provides superior scaling compared to numerically exact solvers, offering a robust and efficient pathway for exploring many-body dissipative quantum dynamics that were previously computationally inaccessible.