Papers.

Predictive Modeling of Opinion and Connectivity Dynamics in Social Networks

Recent years saw an increased interest in modeling and understanding the mechanisms of opinion and innovation spread through human networks. Using analysis of real-world social data, researchers are able to gain a better understanding of the dynamics of social networks and subsequently model the changes in such networks over time. We developed a social network model that both utilizes an agent-based approach with a dynamic update of opinions and connections between agents and reflects opinion propagation and structural changes over time as observed in real-world data. We validate the model using data from the Social Evolution dataset of the MIT Human Dynamics Lab describing changes in friendships and health self-perception in a targeted student population over a nine-month period. We demonstrate the effectiveness of the approach by predicting changes in both opinion spread and connectivity of the network. We also use the model to evaluate how the network parameters, such as the level of `openness' and willingness to incorporate opinions of neighboring agents, affect the outcome. The model not only provides insight into the dynamics of ever changing social networks, but also presents a tool with which one can investigate opinion propagation strategies for networks of various structures and opinion distributions.

Scalable Linear Causal Inference for Irregularly Sampled Time Series with Long Range Dependencies

Dynamic Memory Networks for Visual and Textual Question Answering

Neural network architectures with memory and attention mechanisms exhibit certain reasoning capabilities required for question answering. One such architecture, the dynamic memory network (DMN), obtained high accuracy on a variety of language tasks. However, it was not shown whether the architecture achieves strong results for question answering when supporting facts are not marked during training or whether it could be applied to other modalities such as images. Based on an analysis of the DMN, we propose several improvements to its memory and input modules. Together with these changes we introduce a novel input module for images in order to be able to answer visual questions. Our new DMN+ model improves the state of the art on both the Visual Question Answering dataset and the \babi-10k text question-answering dataset without supporting fact supervision.

MacroBase: Prioritizing Attention in Fast Data

ZeroDB white paper

Sentiment Visualisation Widgets for Exploratory Search

Data Portraits and Intermediary Topics: Encouraging Exploration of Politically Diverse Profiles

Yggdrasil: An Optimized System for Training Deep Decision Trees at Scale

Deep distributed decision trees and tree ensembles have grown in importance due to the need to model increasingly large datasets. However, PLANET, the standard distributed tree learning algorithm implemented in systems such as \xgboost and Spark MLlib, scales poorly as data dimensionality and tree depths grow. We present Yggdrasil, a new distributed tree learning method that outperforms existing methods by up to 24x. Unlike PLANET, Yggdrasil is based on vertical partitioning of the data (i.e., partitioning by feature), along with a set of optimized data structures to reduce the CPU and communication costs of training. Yggdrasil (1) trains directly on compressed data for compressible features and labels; (2) introduces efficient data structures for training on uncompressed data; and (3) minimizes communication between nodes by using sparse bitvectors. Moreover, while PLANET approximates split points through feature binning, Yggdrasil does not require binning, and we analytically characterize the impact of this approximation. We evaluate Yggdrasil against the MNIST 8M dataset and a high-dimensional dataset at Yahoo; for both, Yggdrasil is faster by up to an order of magnitude.

Weighted Log-odds-ratio, Informative Dirichlet Prior Method to Enhance Peer Review Feedback for Low- and High-scoring College Students in a Required First-year Writing Program

Virtual Screening through Substructure Matching

Tutorial: Introduction to Social Media Network Analysis with NodeXL

Towards Streamlined Big Data Analytics

Topic modeling in scientometrics: Community, connectivity, and change

Tools to Use in an Information Technology Class--and Best of All They Are FREE!.

TempWeb 2016 Chairs’ Welcome Message

TASTY Reference Manual

Symmetrical Web Archiving With Webrecorder

Svg

Sparkmania

Solving for High Throughput with Akka Streams

Semantically Annotated Concepts in KDD's 2009-2015 Abstracts

Semantic Knowledge and Privacy in the Physical Web

Scalable Semantic Matching of Queries to Ads in Sponsored Search Advertising

Running Spark on Alluxio with S3

Reactive Microservices Architecture: Design Principles for Distributed Systems

Provable learning of Noisy-or Networks.

Provable Algorithms for Inference in Topic Models.

Programming in Scala: Updated for Scala 2.12

Polynomial-time Tensor Decompositions with Sum-of-Squares.

Persistent RNNs: Stashing Recurrent Weights On-Chip

This paper introduces a new technique for mapping Deep Recurrent Neural Networks (RNN) efficiently onto GPUs by caching recurrent weights in on-chip registers, achieving a 30x speedup over the previous state of the art at low mini-batch sizes and enabling larger, deeper RNNs to be trained efficiently.

Motion Sickness Related Aspects of Inclusion of Color Deficient Observers in Virtual Reality

Motion Sickness Related Aspects of Inclusion of Color Deficient Observers in Virtual Reality

Matrix Completion has No Spurious Local Minimum.

Mastering Scala Machine Learning

MacroBase: Analytic Monitoring for the Internet of Things

Machine Learning with Python and H2O

Linear Algebraic Structure of Word Senses, with Applications to Polysemy.

Interactive Exploration on Large Genomic Datasets

The prevalence of large genomics datasets has made the need to explore this data more important. Large sequencing projects like the 1000 Genomes Project have produced over 200TB of publicly available data, while existing genomic visualization tools have been unable to scale. In this work we present Mango, a scalable genome browser built on top of ADAM that can run both locally and on a cluster, combining optimizations that drive novel genomic visualization techniques over terabytes of genomic data.

Incremental Life Cycle Assurance of Safety-Critical Systems

Finding problems and optimal designs in the requirements phase is more efficient than later phases. However, over-constraining the solution is also sub-optimal since not all information is necessarily available upfront. 'Build-then-test' approaches which insist on developing first requirements, then architecture, then implementation are not suitable for building systems that must be rapidly fielded and respond to ever-changing demands. Our approach, ALISA, is working on integrating four pillars for incrementally building systems which can be shown to satisfy the relevant requirements. Our four key pillars for assuring requirements satisfaction are requirements specifications, architecture models, verification techniques, and assurance case traceability between the first three. In this paper we introduce our approach, and highlight how we are integrating these pillars using an XText-driven DSL and tool meta-model leveraging existing tools and languages. Our current focus is on understanding exactly which requirements are responsible for the majority of design constraints. Identifying this subset promises to reduce architecture design space exploration and verification overhead, increasing delivery cadence.

Identity Matters in Deep Learning.

Heterogeneous Storage in HopsFS

Hard Knocks and Soft Spots: A Docker-Centric CI/CD Pipeline at VMware

Gradient Descent Learns Linear Dynamical Systems.

From Human Language Technology to Human Language Science

Finding Approximate Local Minima for Nonconvex Optimization in Linear Time.

FairRide: Near-Optimal, Fair Cache Sharing

Evaluation of the Australian cybercrime online reporting network

End-to-End Large Scale Machine Learning with KeystoneML

Effective Online and Offline Privacy-Preserving Processing of Location Data

Css3

Cryptocurrencies, blockchains, and smart contracts; Hardware for deep learning