Papers.

OptiML: An Implicitly Parallel Domain-Specific Language for Machine Learning

As the size of datasets continues to grow, machine learning applications are becoming increasingly limited by the amount of available computational power. Taking advantage of modern hardware requires using multiple parallel programming models targeted at different devices (e.g. CPUs and GPUs). However, programming these devices to run efficiently and correctly is difficult, error-prone, and results in software that is harder to read and maintain. We present OptiML, a domain-specific language (DSL) for machine learning. OptiML is an implicitly parallel, expressive and high performance alternative to MATLAB and C++. OptiML performs domain-specific analyses and optimizations and automatically generates CUDA code for GPUs. We show that OptiML outperforms explicitly parallelized MATLAB code in nearly all cases. 1.

On the Design of a Practical Information-Centric Transport

On a Conjecture of Butler and Graham

Numerical Verification of the Power Transfer and Wakefield Coupling in the CLIC Two-Beam Accelerator

The Compact Linear Collider (CLIC) provides a path to a multi-TeV accelerator to explore the energy frontier of High Energy Physics. Its two-beam accelerator (TBA) concept envisions complex 3D structures, which must be modeled to high accuracy so that simulation results can be directly used to prepare CAD drawings for machining. The required simulations include not only the fundamental mode properties of the accelerating structures but also the Power Extraction and Transfer Structure (PETS), as well as the coupling between the two systems. Time-domain simulations will be performed to understand pulse formation, wakefield damping, fundamental power transfer and wakefield coupling in these structures. Applying SLAC's parallel finite element code suite, these large-scale problems will be solved on some of the largest supercomputers available. The results will help to identify potential issues and provide new insights on the design, leading to further improvements on the novel two-beam accelerator scheme.

Networks Analysis of a Regional Ecosystem of Afterschool Programs

Multi-View Learning of Word Embeddings via CCA

NeurRecently, there has been substantial interest in using large amounts of unlabeled data to learn word representations which can then be used as features in supervised classifiers for NLP tasks. However, most current approaches are slow to train, do not model the context of the word, and lack theoretical grounding. In this paper, we present a new learning method, Low Rank Multi-View Learning (LR-MVL) which uses a fast spectral method to estimate low dimensional context-specific word representations from unlabeled data. These representation features can then be used with any supervised learner. LR-MVL is extremely fast, gives guaranteed convergence to a global optimum, is theoretically elegant, and achieves state-ofthe- art performance on named entity recognition (NER) and chunking problems.

Model-Based Engineering Approach for System Architecture Exploration

MobileWorks: A Mobile Crowdsourcing Platform for Workers at the Bottom of the Pyramid

MIREX 2011 AUDIO TAG CLASSIFICATION USING WEIGHTED-VOTE NEAREST NEIGHBOR CLASSIFICATION

Mind Economy: Dynamic Graph Analysis of Communications

Micro-scheduling and its interaction withcache partitioning

Mesos: Flexible Resource Sharing for the Cloud

Mesos: A Platform for Fine-Grained Resource Sharing in the Data Center

Machine Learning in the Cognitive Foundry.

Lift in Action

Large-Scale Cross-Document Coreference Using Distributed Inference and Hierarchical Models

Large-Scale Cognate Recovery

Kafka : a Distributed Messaging System for Log Processing

Investigating Factors Influencing Crowdsourcing Tasks with High Imaginative Load

Interoperability of Registries

International Temporal Web Analytics Workshop TWAW 2011

Inspecting Sample Reusability for Active Learning.

How Can an Improved Business Environment Contribute to Export-led Growth in Serbia?

Generic Proof Tools and Finite Group Theory

Functor is to Lens as Applicative is to Biplate: Introducing Multiplate

Functional Programming for Java Developers

Frontier Pruning for Shift-Reduce CCG Parsing

Fast and Balanced: Efficient Label Tree Learning for Large Scale Object Recognition

Experiences with JSON and XML Transformations

Entracker : Energy Tracker for Homes

Dynamic Pooling and Unfolding Recursive Autoencoders for Paraphrase Detection

Drug knowledge expressed as computable semantic triples

Dominant Resource Fairness: Fair Allocation of Multiple Resource Types

Crisis MT: Developing A Cookbook for MT in Crisis Situations

Conflux: Embedding Massively Parallel Semantics in a High-Level Programming Language

Cognitive policy learner: biasing winning or losing strategies

Automatic Measurement and Comparison of Vowel Nasalization across Languages

Approximate Nearest Neighbor Search Small World Approach

Analyzing the Dynamics of Research by Extracting Key Aspects of Scientific Papers

Advances in Parallel Electromagnetic Codes for Accelerator Science and Development

Over a decade of concerted effort in code development for accelerator applications has resulted in a new set of electromagnetic codes which are based on higher-order finite elements for superior geometry fidelity and better solution accuracy. SLAC's ACE3P code suite is designed to harness the power of massively parallel computers to tackle large complex problems with the increased memory and solve them at greater speed. The US DOE supports the computational science R&D under the SciDAC project to improve the scalability of ACE3P, and provides the high performance computing resources needed for the applications. This paper summarizes the advances in the ACE3P set of codes, explains the capabilities of the modules, and presents results from selected applications covering a range of problems in accelerator science and development important to the Office of Science.

A Study of Boosting based Transfer Learning for Activity and Gesture Recognition

Real-world environments are characterized by non-stationary and continuously evolving data. In this thesis, a novel instance transfer technique that adapts a 'Cost-sensitive' variation of AdaBoost is presented. The method capitalizes on the theoretical and functional properties of AdaBoost to selectively reuse outdated training instances obtained from a 'source' domain to effectively classify unseen instances occurring in a different, but related 'target' domain. The algorithm is evaluated on accelerometer based 3D gesture recognition, smart home activity recognition and text categorization.

A Semantic Model for Integrated Content Management, Localisation and Language Technology Processing

A Reduced Gradient Output Design for SLAC's XL4 X-Band Klystron

A recommendation engine for Wikipedia articles based on constrained training data

A novel role for ITAM-containing proteins in 2B4-initiated signaling and cytotoxicity

A New Variation of Hat Guessing Games

50 MW X-Band RF System for a Photoinjector Test Station at LLNL

Abstract Not Provided

Resigned to quitting

Adapting to Trends in Language Resource Development: A Progress Report on LDC Activities

Concept Mentions within KDD-2009 Abstracts (kdd09cma1) Linked to a KDD Ontology (kddo1)

We introduce the kddo1 ontology and semantically annotated kdd09cma1 corpus from the field of knowledge discovery in database (KDD) research. The corpus is based on the abstracts for the papers accepted into the KDD-2009 conference. Each abstract has its concept mentions identified and, where possible, linked to the appropriate concept in the ontology. The ontology is based on a human generated and readable semantic wiki focused on concepts and relationships for the domain along with other related topics, papers and researchers from information sciences. To our knowledge this is the first ontology and interlinked corpus for a subdiscipline within computing science. The dataset enables the evaluation of supervised approaches to semantic annotation of documents that contain a large number of high-level concepts relative the number of named entity mentions. We plan to continue to evolve the ontology based on the discovered relations within the corpus and to extend the corpus to cover other research paper abstracts from the domain. Both resources are publicly available at http://www.gabormelli.com/Projects/kdd/data/.