Stochastic Gradient Boosted Decision Trees (GBDT) is one of the most widely used learning algorithms in machine learning today. It is adaptable, easy to interpret, and produces highly accurate models. However, most implementations today are computationally expensive and require all training data to be in main memory. As training data becomes ever larger, there is motivation for us to parallelize the GBDT algorithm. Parallelizing decision tree training is intuitive and various approaches have been explored in existing literature. Stochastic boosting on the other hand is inherently a sequential process and have not been applied to distributed decision trees. In this work, we present two different distributed methods that generates exact stochastic GBDT models, the first is a MapReduce implementation and the second utilizes MPI on the Hadoop grid environment.

Enterprise and web data processing and content aggregation systems often require extensive use of human-reviewed data (e.g. for training and monitoring machine learning-based applications). Today these needs are often met by in-house efforts or out-sourced offshore contracting. Emerging applications attempt to provide automated collection of human-reviewed data at Internet-scale. We conduct extensive experiments to study the effectiveness of one such application. We also study the feasibility of using Yahoo! Answers, a general question-answering forum, for human-reviewed data collection.

HTML has so far been the standard format for delivering information on the World Wide Web. However, automated information processing of these documents for data exchange and interoperability has been difficult. XML, a subset of SGML, has been proposed to be the next standard format, that allows user-defined tags for better describing nested document structures and associated semantics. Operations on structured documents, such as searching in nested document structures, require new functions that are not currently available on most systems today. We describe a general framework for manipulating structured documents based on document abstractions. An abstraction is an approximation of an actual document, while possessing useful properties for analyses of interest. The framework provides a wide design space for the tradeoff between cost and capability. This general framework can be applied to index design, document searching and categorization. We present this framework by focusing on the indexing and searching of structured documents in the XML domain, and prove their soundness. We also address the issue of rich data types in XML documents.

False sharing can be a source of significant overhead on shared-memory multiprocessors. Several program restructuring techniques to reduce false sharing have been proposed in past work. In this paper, we propose an approach for elimination of false sharing based solely on selection of runtime schedule parameters for parallel loops. This approach leads to more portable code since only the schedule parameters need to be changed to target different multiprocessors. Also, the guarantee of elimination (rather than reduction) of false sharing in a parallel loop can significantly reduce the bookkeeping overhead in some memory consistency mechanisms. We present some preliminary experimental results for this approach.

Traditional optimization techniques for sequential programs are not directly applicable to parallel programs where concurrent activities may interfere with each other through shared variables. New compiler techniques must be developed to accommodate features found in parallel languages. In this paper, we use abstract interpretation to obtain useful properties of programs, e.g., side effects, data dependences, object lifetime and concurrent expressions, for a language that supports first-class functions, pointers, dynamic allocations and explicit parallelism through cobegin. These analyses may facilitate many applications, such as program optimization, parallelization, restructuring, memory management, and detecting access anomalies.

The authors discuss run-time microtasking support for executing nested parallel loops on a shared-memory multiprocessor system, and present a scheme called switch-stacks for implementing such support. They first discuss current approaches to flat microtasking, and investigate how to extend these to full microtasking. They point out the problem of dummy waiting in the processor that initiates a parallel loop. To solve this problem, two schemes, dequeue-tasks and dequeue-descendant-tasks, are considered, and their disadvantages are discussed. The proposed switch-stack scheme perfectly solves the problem. These schemes have been implemented in the PARCEL run-time system. The results show that the new scheme nearly always achieves the best performance in execution time and stability.>