Spark (mathematics)

About: Spark (mathematics) is a research topic. Over the lifetime, 7304 publications have been published within this topic receiving 63322 citations.

A Parallelizing Approach to the High-Level Synthesis of Digital Circuits

Abstract: SPARK: A Parallelizing High – Level Synthesis of Digital Circuits presents a novel approach to the high-level synthesis of digital circuits – that of parallelizing high-level synthesis (PHLS). This approach uses aggressive code parallelizing and code motion techniques to discover circuit optimization opportunities beyond what is possible with traditional high-level synthesis. This PHLS approach addresses the problems of the poor quality of synthesis results and the lack of controllability over the transformations applied during the high-level synthesis of system descriptions with complex control flows, that is, with nested conditionals and loops.

A Study on the Use of Combustion Phase Indicators for MBT Spark Timing on a Bi-Fuel Engine

Abstract: The performance of a spark ignition engine strongly depends on the phase of the combustion process with respect to piston motion, and hence on the spark advance; this fundamental parameter is actually controlled in open-loop by means of maps drawn up on the test bench and stored in the Electronic Control Unit (ECU). Bi-fuel engines (e.g. running either on gasoline or on natural gas) require a double mapping process in order to obtain a spark timing map for each of the fuels. This map based open-loop control however does not assure to run the engine always with the best spark timing, which can be influenced by many factors, like ambient condition of pressure, temperature and humidity, fuel properties, engine wear. A feedback control instead can maintain the spark advance at its optimal value apart from operative and boundary conditions, so as to gain the best performance (or minimum fuel consumption). Such a control can be realized using as pilot variable a combustion phase indicator, i.e. a parameter which depends exclusively on the phase of the heat release process and assumes a fixed value for optimal spark timing. The purpose of the present work is to compare the behaviour of the most used combustion phase indicators using two different fuels one after the other (common gasoline and Compressed Natural Gas, CNG) on the same engine, in order to assess the influence of different heat release progress and to verify the possibility to feedback control the spark timing apart from the fuel used. The comparison has been carried on by means of experimental test on the engine test bench, analysing incylinder pressure acquired with varying spark advance for different operative conditions of engine speed, load and air-to-fuel ratio.

Low-inductance multigap spark modules

Abstract: We outline the design concept for low-inductance high-current spark modules at a voltage level of 100 kV and a current of 1 MA. We present the results of an investigation of the switching and operating characteristics of multichannel, multigap spark modules as a function of the design and the shape and amplitude of the beam pulse. We give a description of the designs and parameters of the developed types of spark modules.

In-cylinder boosting of turbocharged spark-ignited engines. Part 1: Model-based design of the charge valve

Abstract: Downsizing and turbocharging for retaining the maximal power is a widely used approach to decrease the fuel consumption of spark ignited engines. In general, the trade-off is a substantial driveability loss. In-cylinder boosting has proven to be an effective way to eliminate this problem. Thus far, expensive and complex fully variable valve-trains have been proposed for the air exchange between the air tank and the combustion chamber. This paper is the first of a two-part study that examines the use of a deactivatable camshaft-driven valve with respect to the achievable transient engine performance. The system characteristics and limitations are discussed by using a mean value engine model that is adapted for in-cylinder boosting. A model-based design framework is presented which links the valve system design to a desired engine performance. The companion paper covers control issues and provides experimental verifications.

Benchmarking Data Flow Systems for Scalable Machine Learning

Abstract: Distributed data flow systems such as Apache Spark or Apache Flink are popular choices for scaling machine learning algorithms in production. Industry applications of large scale machine learning such as click-through rate prediction rely on models trained on billions of data points which are both highly sparse and high-dimensional. Existing Benchmarks attempt to assess the performance of data flow systems such as Apache Flink, Spark or Hadoop with non-representative workloads such as WordCount, Grep or Sort. They only evaluate scalability with respect to data set size and fail to address the crucial requirement of handling high dimensional data.We introduce a representative set of distributed machine learning algorithms suitable for large scale distributed settings which have close resemblance to industry-relevant applications and provide generalizable insights into system performance. We implement mathematically equivalent versions of these algorithms in Apache Flink and Apache Spark, tune relevant system parameters and run a comprehensive set of experiments to assess their scalability with respect to both: data set size and dimensionality of the data. We evaluate the systems for data up to four billion data points and 100 million dimensions. Additionally we compare the performance to single-node implementations to put the scalability results into perspective.Our results indicate that while being able to robustly scale with increasing data set sizes, current state of the art data flow systems are surprisingly inefficient at coping with high dimensional data, which is a crucial requirement for large scale machine learning algorithms.