Abstract
Reducing computational scaling for simulating non-Markovian dissipativedynamics using artificial neural networks is both a major focus and formidablechallenge in open quantum systems. To enable neural quantum states (NQSs), weencode environmental memory in dissipatons (quasiparticles with characteristiclifetimes), yielding the dissipaton-embedded quantum master equation (DQME).The resulting NQS-DQME framework achieves compact representation of many-bodycorrelations and non-Markovian memory. Benchmarking against numerically exacthierarchical equations of motion confirms NQS-DQME maintains comparableaccuracy while enhancing scalability and interpretability. This methodologyopens new paths to explore non-Markovian open quantum dynamics in previouslyintractable systems.