Data Management in Networks: Experimental Evaluation of a Provably Good Strategy

This paper deals with data management for parallel and distributed systems. We present the DIVA (Distributed Variables ) library that provides direct access to shared data objects from each node in a network. The current implementations are based on mesh-connected massively parallel computers. Our algorithms dynamically create and discard copies of the data objects in order to reduce the communication overhead. We use a non-standard approach based on a randomized but locality preserving embedding of ``access trees' into the network. The access tree strategy was previously analyzed only in a theoretical model. A competitive analysis proved that the strategy minimizes the network congestion up to small factors. In this paper the access tree strategy is evaluated experimentally. We test several variations of this strategy on three different applications of parallel computing, namely matrix multiplication, bitonic sorting, and Barnes—Hut N -body simulation. We compare the access tree strategy with a standard caching strategy using a fixed home for each data object. Our experiments show that the access tree strategy outperforms the fixed home strategy clearly as it avoids network congestion. Furthermore, we do comparisons with hand-optimized message passing strategies. In fact, we observe that the access tree strategy comes reasonably close to the performance of the hand-optimized message passing strategies while the fixed home strategy performs poorly. Received in final form October 8, 2001. Online publication March 25, 2002.     

Trial and error

A pac-learning algorithm isd-space bounded, if it stores at mostd examples from the sample at any time. We characterize thed-space learnable concept classes. For this purpose we introduce the compression parameter of a concept classb and design our Trial and Error Learning Algorithm. We show: b isd-space learnable if and only if the compression parameter ofb is at mostd. This learning algorithm does not produce a hypothesis consistent with the whole sample as previous approaches e.g. by Floyd, who presents consistent space bounded learning algorithms, but has to restrict herself to very special concept classes. On the other hand our algorithm needs large samples; the compression parameter appears as exponent in the sample size. We present several examples of polynomial time space bounded learnable concept classes:

- all intersection closed concept classes with finite VC-dimension.

- convexn-gons in ?².

- halfspaces in ?ⁿ.

- unions of triangles in ?².

We further relate the compression parameter to the VC-dimension, and discuss variants of this parameter.     

Strongly adaptive token distribution

The token distribution (TD) problem, an abstract static variant of load balancing, is defined as follows: letM be a (parallel processor) network with setP of processors. Initially, each processorP P has a certain amountl(P) of tokens. The goal of a TD algorithm, run onM, is to distribute the tokens evenly among the processors. In this paper we introduce the notion of strongly adaptive TD algorithms, i.e., algorithms whose running times come close to the best possible runtime, the off-line complexity of the TD problem, for each individual initial token distributionl. Until now, only weakly adaptive algorithms have been considered, where the running time is measured in terms of the maximum initial load max{l(P)P P}.We design an almost optimal, strongly adaptive algorithm on mesh-connected networks of arbitrary dimension extended by a single 1-bit bus. This result shows that an on-line TD algorithm can come close to the optimal (off-line) bound for each individual initial load. Furthermore, we exactly characterize the off-line complexity of arbitrary initial token distributions on arbitrary networks. As an intermediate result, we design almost optimal weakly adaptive algorithms for TD on mesh-connected networks of arbitrary dimension.This research was partially supported by DFG-Forschergruppe Effiziente Nutzung massiv paralleler Systeme, Teilprojekt 4, by the ESPRIT Basic Research Action No. 7141 (ALCOM II), and by the Volkswagen-stiftung. A preliminary version was presented at the 20th ICALP, 1993, see [9].     

Efficient PRAM simulation on a distributed memory machine

We present algorithms for the randomized simulation of a shared memory machine (PRAM) on a Distributed Memory Machine (DMM). In a PRAM, memory conflicts occur only through concurrent access to the same cell, whereas the memory of a DMM is divided into modules, one for each processor, and concurrent accesses to the same module create a conflict. Thedelay of a simulation is the time needed to simulate a parallel memory access of the PRAM. Any general simulation of anm processor PRAM on ann processor DMM will necessarily have delay at leastm/n. A randomized simulation is calledtime-processor optimal if the delay isO(m/n) with high probability. Using a novel simulation scheme based on hashing we obtain a time-processor optimal simulation with delayO(log log(n) log*(n)). The best previous simulations use a simpler scheme based on hashing and have much larger delay: (log(n)/log log(n)) for the simulation of an n processor PRAM on ann processor DMM, and (log(n)) in the case where the simulation is time-processor optimal.Our simulations use several (two or three) hash functions to distribute the shared memory among the memory modules of the PRAM. The stochastic processes modeling the behavior of our algorithms and their analyses based on powerful classes of universal hash functions may be of independent interest.Research partially supported by NSF/DARPA Grant CCR-9005448. Work was done while at the University of California at Berkeley and the International Computer Science Institute, Berkeley, CA.Research partially supported by National Science Foundation Operating Grant CCR-9016468, National Science Foundation Operating Grant CCR-9304722, United States-Israel Binational Science Foundation Grant No. 89-00312, United States-Israel Binational Science Foundation Grant No. 92-00226, and ESPRIT BR Grant EC-US 030.Part of work was done during a visit at the International Computer Science Institute at Berkeley; supported in part by DFG-Forschergruppe Effiziente Nutzung massiv paralleler Systeme, Teilprojekt 4, and by the Esprit Basic Research Action Nr. 7141 (ALCOM II).     

Cloud Computing Based Optimal Driving for a Parallel Hybrid Electric Vehicle

A cloud computing based optimal driving method is proposed and its feasibility is validated through a real-world scenario simulation. Based on principles of vehicle dynamics, the driving optimization problem has been formulated into an optimal control problem constrained by traffic rules, directed at achieving lower equivalent fuel consumption and shorter travel time. In order to conveniently specify the constraints and facilitate the application of the dynamic programming (DP) algorithm, the driving optimization problem is transformed into spatial domain and discretized properly. Considering the heavy computational costs of the DP algorithm, a cloud computing based platform structure is proposed to solve the optimal driving problem in real-time. A case study is simulated based on a real-world traffic scenario in Matlab. Simulation results demonstrate that the cloud computing framework is promising toward realizing the real-time energy management for hybrid electric vehicles.     

Attention training in individuals with generalized social phobia: A randomized controlled trial.

The authors conducted a randomized, double-blind placebo-controlled trial to examine the efficacy of an attention training procedure in reducing symptoms of social anxiety in 44 individuals diagnosed with generalized social phobia (GSP). Attention training comprised a probe detection task in which pictures of faces with either a threatening or neutral emotional expression cued different locations on the computer screen. In the attention modification program (AMP), participants responded to a probe that always followed neutral faces when paired with a threatening face, thereby directing attention away from threat. In the attention control condition (ACC), the probe appeared with equal frequency in the position of the threatening and neutral faces. Results revealed that the AMP facilitated attention disengagement from threat from pre- to postassessment and reduced clinician- and self-reported symptoms of social anxiety relative to the ACC. The percentage of participants no longer meeting Diagnostic and Statistical Manual (4th ed.) criteria for GSP at postassessment was 50% in the AMP and 14% in the ACC. Symptom reduction in the AMP group was maintained during 4-month follow-up assessment. These results suggest that computerized attention training procedures may be beneficial for treating social phobia. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Influence of measurement procedure on quality of impedance spectra on lead-acid batteries

Many battery simulation models, but also electrochemical interpretations are based on impedance spectroscopy. However, the impedance of a battery is influenced by various factors, e.g. in the case of a lead-acid battery: state of charge (SOC), charging or discharging, superimposed dc current, short-term history or homogeneity of the electrolyte. This paper analyses the impact of those factors on impedance spectra of lead-acid batteries. The results show that very detailed information about the conditions during the measurement is crucial for the correct interpretation of a spectrum.     

Epoxide–tertiary amine combinations as efficient catalysts for methanol carbonylation into methyl formate in the presence of carbon dioxide

The influence of CO₂ on the carbonylation of methanol into methyl formate was investigated with two classes of catalytic systems: MeO⁻ and epoxide–amine combinations. In the absence of CO₂, NaOMe is the more active, but its efficiency is rapidly destroyed by CO₂. On the other hand, the epoxide–amine system is much less sensitive to the presence of CO₂. Different classes of amines and epoxides have been tested. Tertiary alkyl amines and terminal epoxides are the most efficient. Phosphines associated with an epoxide have also been found to be able to catalyse this reaction. On the basis of

Full-size image

NMR studies, a plausible mechanism which can also be applied to amines is proposed for the reaction.     

A Convolution Algebra of Delay-Differential Operators and a Related Problem of Finite Spectrum Assignability

In this paper we investigate a ring of delay-differential operators, which appears in the study of time-delay systems with commensurate point delays. Besides point-delay differential operators, this ring also contains certain distributed delays. It can be described as a certain algebraic extension of the ring of point-delay differential operators as well as a convolution algebra of compact support distributions. It will be shown that the finite spectrum assignability problem can be solved for spectrally controllable systems with controllers from this algebra. Date received: February 24, 1998. Date revised: February 9, 1999.