Relativistic hydrodynamics is a powerful tool to simulate the evolution of the quark gluon plasma (QGP) in relativistic heavy ion collisions. Using 10000 initial and final profiles generated from 2+1-d relativistic hydrodynamics VISH2+1 with MC-Glauber initial conditions, we train a deep neural network based on stacked U-net, and use it to predict the final profiles associated with various initial conditions, including MC-Glauber, MC-KLN and AMPT and TRENTo. A comparison with the VISH2+1 results shows that the network predictions can nicely capture the magnitude and inhomogeneous structures of the final profiles, and nicely describe the related eccentricity distributions $P(\varepsilon_n)$ (n=2, 3, 4). These results indicate that deep learning technique can capture the main features of the non-linear evolution of hydrodynamics, showing its potential to largely accelerate the event-by-event simulations of relativistic hydrodynamics.

The long-term, highly periodic and correlated variations in both the pulse shape and the rate of slow-down of two isolated pulsars (PSRs) PSR B1828-11 and PSR B1642-03 were discovered recently. This phenomenon may provide evidence for "free precession" as suggested in the literature. Some authors presented various kinds of models to explain this phenomenon within the framework of free precession. Here we present an accretion disk model for this precession phenomenon instead. Under reasonable parameters, the observed phenomenon can be explained by an isolated pulsar with a fossil disk. This may link radio pulsars and anomalous X-ray pulsars (AXPs) and present an indirect evidence for the existence of the fossil disk in nature.