All issues

А new cloud center of parallel computing is to be created in the Laboratory of Information Technologies (LIT) of the Joint Institute for Nuclear Research JINR) what is expected to improve significantly the efficiency of numerical calculations and expedite the receipt of new physically meaningful results due to the more rational use of computing resources. To optimize a scheme of parallel computations at a cloud environment it is necessary to test this scheme for various combinations of equipment parameters (processor speed and numbers, throughput оf а communication network etc). As a test problem, the parallel MPI algorithm for calculations of the long Josephson junctions (LDJ) is chosen. Problems of evaluating the impact of abovementioned factors of computing mean on the computing speed of the test problem are solved by simulation with the simulation program SyMSim developed in LIT.

The simulation of the LDJ calculations in the cloud environment enable users without a series of test to find the optimal number of CPUs with a certain type of network run the calculations in a real computer environment. This can save significant computational time in countable resources. The main parameters of the model were obtained from the results of the computational experiment conducted on a special cloud-based testbed. Computational experiments showed that the pure computation time decreases in inverse proportion to the number of processors, but depends significantly on network bandwidth. Comparison of results obtained empirically with the results of simulation showed that the simulation model correctly simulates the parallel calculations performed using the MPI-technology. Besides it confirms our recommendation: for fast calculations of this type it is needed to increase both, — the number of CPUs and the network throughput at the same time. The simulation results allow also to invent an empirical analytical formula expressing the dependence of calculation time by the number of processors for a fixed system configuration. The obtained formula can be applied to other similar studies, but requires additional tests to determine the values of variables.

This article reviews both historical achievements and modern results in the field of Markov Decision Process (MDP) and convex optimization. This review is the first attempt to cover the field of reinforcement learning in Russian in the context of convex optimization. The fundamental Bellman equation and the criteria of optimality of policy — strategies based on it, which make decisions based on the known state of the environment at the moment, are considered. The main iterative algorithms of policy optimization based on the solution of the Bellman equations are also considered. An important section of this article was the consideration of an alternative to the $Q$-learning approach — the method of direct maximization of the agent’s average reward for the chosen strategy from interaction with the environment. Thus, the solution of this convex optimization problem can be represented as a linear programming problem. The paper demonstrates how the convex optimization apparatus is used to solve the problem of Reinforcement Learning (RL). In particular, it is shown how the concept of strong duality allows us to naturally modify the formulation of the RL problem, showing the equivalence between maximizing the agent’s reward and finding his optimal strategy. The paper also discusses the complexity of MDP optimization with respect to the number of state–action–reward triples obtained as a result of interaction with the environment. The optimal limits of the MDP solution complexity are presented in the case of an ergodic process with an infinite horizon, as well as in the case of a non-stationary process with a finite horizon, which can be restarted several times in a row or immediately run in parallel in several threads. The review also reviews the latest results on reducing the gap between the lower and upper estimates of the complexity of MDP optimization with average remuneration (Averaged MDP, AMDP). In conclusion, the real-valued parametrization of agent policy and a class of gradient optimization methods through maximizing the $Q$-function of value are considered. In particular, a special class of MDPs with restrictions on the value of policy (Constrained Markov Decision Process, CMDP) is presented, for which a general direct-dual approach to optimization with strong duality is proposed.

This work is devoted to molecular dynamic modeling of the thermal impact processes on the metal sample consisting of nickel atoms. For the solution of this problem, a continuous mathematical model on the basis of the classical Newton mechanics equations has been used; a numerical method based on the Verlet scheme has been chosen; a parallel algorithm has been offered, and its realization within the MPI and OpenMP technologies has been executed. By means of the developed parallel program, the investigation of thermodynamic equilibrium of nickel atoms’ system under the conditions of heating a sample to desired temperature has been executed. In numerical experiments both optimum parameters of calculation procedure and physical parameters of analyzed process have been defined. The obtained numerical results are well corresponding to known theoretical and experimental data.

The paper presents a software system aimed at finding best (in some sense) parameters of an algorithm. The system handles both discrete and continuous parameters and employs massive parallelism offered by public clouds. The paper presents an overview of the system, a method to measure algorithm's performance in the cloud and numerical results of system's use on several problem sets.