Search

The paper describes an approach to parallelization of Normal Bayes classifier training algorithm for distributed data. In the process of distributed data analysis and the algorithm performance, the results fail to join properly. Due to this, the algorithm is to be performed in a distributed manner. For this purpose, we use representation of the algorithm as a sequential composition of functions. The algorithm is parallelized to work with data distributed horizontally and vertically. This allows placing parallel functions of the algorithm at data nodes. Experiments show that transfer of computations to sources allow to decrease training time and network traffic. We implement the algorithm variants as an extension of the industrial-strength Java-based library Xelopes.

A very high-current liner implosion experiment was conducted, using an explosive magnetic-compression generator (EMG) to deliver a peak current of 102 /spl plusmn/ 3 MA, to implode a 4.0-mm-thick aluminum liner. Analysis of experimental data showed that the inner surface of the liner had attained a velocity of between 6.8-8.4 km/s, consistent with detailed numerical calculations. Both calculations and data were consistent with a final liner state that was still substantially solid at target impact time and had a total kinetic energy of over 20 MJ.

In the three joint VNIIEF/LANL experiments RHSR-0,1,2 with a disk EMG and a three-layer liner system (Al-dielectric-Cu) a method of radial radiography was used to study growth of perturbations amplitude at the boundary between copper layer and polyethylene (in experiments RHSR-0,1) and water (in experiment RHSR-2); the growth occurs during the evolution of the Raleigh-Taylor instability of this boundary.

The imploding liner is a cylinder of conducting material through which a current is passed in the longitudinal direction. Interaction of the current with its own magnetic field causes the liner to implode. In August, 1996, a high energy liner experiment (HEL-1) was conducted at the All-Russia Scientific Research Institute (VNIIEF) in Sarov, Russia. A 5 tier 1 meter diameter explosive disk generator provided electrical energy to drive a 48 cm outside diameter, 4 mm thick, aluminum alloy liner having a mass of about 1 kg onto an 11 cm diameter diagnostic package. The purpose of the experiment was to measure performance of the explosive pulse power generator and the heavy imploding liner. Electrical performance diagnostics included inductive (B-dot) probes, Faraday rotation current measurement, Rogowski total current measurement, and voltage probes, flux loss and conductor motion diagnostics included current-joint voltage measurements and motion sensing contact pins. Optical and electrical impact pins, inductive (B-dot) probes, manganin pressure probes, and continuously recording resistance probes in the central measuring unit (CMU) and piezo and manganin pressure probes, optical beam breakers, and inductive probes located in the glide planes were used as liner symmetry and velocity diagnostics. Preliminary analysis of the data indicate that a peak current of more than 100 MA was attained and the liner velocity was between 6.7 km/sec and 7.5 km/sec. Liner kinetic energy was between 22 MJ and 35 MJ.

The paper describes arrangement and the results of the first joint experiment between VNIIEF and LANL with explosive magnetic generators (EMG) of 1 m diameter and a nonevaporating liner. The experiment took place in August 22, 1996. The goal of the experiment was to accelerate magnetically cylindrical relatively thin aluminum liner and to get kinetic energy of 20 MJ or more. As the energy source for the experimental device we chose 5-module DEMG of 1000 mm diameter, tested many times in the experiments for rigid and liner loads. This EMG can store more energy than any other EMG created at VNIIEF. The physical scene of the liner unit was chosen so that the growth of disturbances would have less influence on the liner shape during flight, especially on the liner's inner surface. The shape of glade plane on which the liner slides during flight and the way of contact liner with walls were chosen on the grounds of 2-D calculations, proceeding from the necessity to ensure electrical contact during the liner flight.

In August 1997, the Los Alamos National Laboratory (LANL) and the All-Russian Scientific Research Institute of Experimental Physics (VNIIEF) conducted a joint experiment in Sarov, Russia to demonstrate the feasibility of applying explosive pulsed power technology to implode large scale, high velocity cylindrical liners. Kilogram mass metal liners imploding at velocities of 5-25 km/sec are useful scientific tools for producing high energy density environments, ultra-high pressure shocks and for the rapid compression of plasmas. To explore the issues associated with the design, operation and diagnosis of such implosions, VNIIEF and LANL designed and executed a practical demonstration experiment in which a liner of approximately 1 kg mass was accelerated to 5-10 km/sec while undergoing a convergence of about 4:1. The scientific objectives of the experiment were three-fold: first to explore the limits of very large, explosive, pulse power systems delivering about 100 MA as drivers for accelerating solid density imploding liners to kinetic energies of 25 MJ or greater; second to evaluate the behavior of single material (aluminum) liners imploding at 5-10 km/s velocities by comparing experimental data with 1-D and 2 D numerical simulations; and third, to evaluate the condition of the selected liner at radial convergence of 4 and a final radius of 6 cm. A liner of such parameters could be used as a driver for the equation of state measurements at megabar pressures or as a driver for a future experiment in which a magnetized fusion plasma would be compressed to approach ignition conditions.