Neural-network variational quantum algorithm for simulating many-body dynamics

We propose a neural-network variational quantum algorithm to simulate the time evolution of many-body systems. Based on modified restricted Boltzmann machine (RBM) wavefunction ansatz, proposed can be efficiently implemented in near-term computers with low measurement cost. Using qubit recycling strategy, only one ancilla is required represent all hidden spins an RBM architecture. The extended open systems by employing stochastic Schrodinger equation approach. Numerical simulations spin-lattice models demonstrate that our capable capturing dynamics closed and high accuracy without suffering from vanishing gradient (or 'barren plateau') issue for considered system sizes.