1: \begin{abstract}

2: The near-critical unitary dynamics of quantum Ising spin chains in transversal and longitudinal magnetic fields is studied using an artificial neural network representation of the wave function.

3: A focus is set on strong spatial correlations which build up in the system following a quench into the vicinity of the quantum critical point.

4: We compare correlations observed following reinforcement learning of the network states with analytical solutions in integrable cases and tDMRG simulations, as well as with predictions from a semi-classical discrete Truncated Wigner analysis. 

5: While the semi-classical approach excells mainly at short times and for small transverse fields, the neural-network representation provides accurate results for a much wider range of parameters. 

6: Where long-range spin-spin correlations build up in the long-time dynamics we find qualitative agreement with exact results while quantitative deviations are of similar size as for the semi-classically predicted correlations, and slow convergence is observed when increasing the number of hidden neurons.

7: \end{abstract}

8: