Convolutional Neural Nets, which is a powerful method of Deep Learning, is applied to classify equation of state of heavy-ion collision event generated within the UrQMD model. Event-by-event transverse momentum and azimuthal angle distributions of protons are used to train a classifier. An overall accuracy of classification of 98% is reached for Au+Au events at √ s N N = 11 GeV. Performance of classifiers, trained on events at different colliding energies, is investigated. Obtained results indicate extensive possibilities of application of Deep Learning methods to other problems in physics of heavyion collisions.

Basic features of directed and elliptic flows of identified hadrons in heavy-ion collisions at intermediate and high energies are considered within two transport string models, UrQMD and QGSM. Both models indicate changing of the sign of proton directed flow at midrapidity from antiflow to normal flow with decreasing energy of collisions. The origin of this effect is traced to hadron rescattering in baryon-rich remnants of the colliding nuclei. To distinguish the effect of rescattering from the flow softening caused by creation of quark-gluon plasma one has to compare heavy-ion and light-ion collisions at the same energy. Both directed and elliptic flows at midrapidity are formed within t = 10-12 fm/c. The differences in the development of elliptic flows of mesons and baryons are found at high energies. These differences can be explained by dissimilar freeze-out conditions, thus suggesting simultaneous study of particle collective flow and freeze-out.

Directed flow of charged particles is studied in nucleus-nucleus collisions simulated within the energy range accessible for NICA and FAIR facilities. Two transport cascade models, UrQMD and QGSM, are employed. These models use different mechanisms of the string excitation and string fragmentation. Despite of the differences, directed flows of charged pions and charged kaons in both models remain antiflow-oriented with reduction of the collision energy from √ s = 11.5 GeV to 3.5 GeV. In contrast, the directed flow of protons changes its sign from antiflow to normal flow within the investigated energy interval. Both models favor continuous non-uniform emittence of hadrons from the expanding fireball rather than sharp, or sudden, freeze-out picture adopted by majority of hydrodynamic models. We found that the earlier frozen hadrons carry the strongest directed flow at midrapidity, although the flow development even at |y| ≤ 0.5 takes about 8-12 fm/c for different hadron species.