Papers.

Optimizing invokedynamic

BTWorld

An energy efficient cache design using spin torque transfer (STT) RAM

The on-chip memory is a dominant source of power and energy consumption in modern and future processors. This paper explores the use of a new emerging non-volatile memory technology as a replacement for SRAM based lower level caches - Spin Torque Transfer(STT) RAM. While STTRAM achieves a reduction in leakage energy of 90% compared to SRAM, the dynamic energy for a write operation is 2X that of SRAM. Consequently, we propose additional microarchitectural optimizations to reduce overall dynamic energy which achieve an average reduction in dynamic energy over the base case of 30% with a range of 16% to 60% across 10 benchmarks.

BibBase triplified

Exploiting scale invariant dynamics for efficient information propagation in large teams

Large heterogeneous teams will often be in situations where sensor data that is uncertain and conflicting is shared across a peer-to-peer network. Not every team member will have direct access to sensors and team members will be influenced mostly by teammates with whom they communicate directly. In this paper, we investigate the dynamics and emergent behaviors of a large team sharing beliefs to reach conclusions about the world. We find empirically that the dynamics of information propagation in such belief sharing systems are characterized by information avalanches of belief changes caused by a single additional sensor reading. The distribution of the size of these avalanches dictates the speed and accuracy with which the team reaches conclusions. A key property of the system is that it exhibits qualitatively different dynamics and system performance over small changes in system parameter ranges. In one particular range, the system exhibits behavior known as scale-invariant dynamics which we empirically find to correspond to dramatically more accurate conclusions being reached by team members. Due to the fact that the ranges are very sensitive to configuration details, the parameter ranges over which specific system dynamics occur are extremely difficult to predict precisely. In this paper we (a) develop techniques to mathematically characterize the dynamics of the team belief propagation (b) obtain through simulations the relation between the dynamics and overall system performance, and (c) develop a novel distributed algorithms that the agents in the team use locally to steer the whole team to areas of optimized performance.

Web retrieval

Query forwarding in geographically distributed search engines

Automatic summarization of cricket video events using genetic algorithm

Network quantification despite biased labels

Quilt: A Patchwork of Multicast Regions

Subject metadata support powered by Maui

P2P VoD using the self-organizing gradient overlay network

Enterprise information extraction: recent developments and open challenges

QRelX: generating meaningful queries that provide cardinality assurance

Privacy-preserving query log mining for business confidentiality protection

Composing parallel software efficiently with lithe

Boolean satisfiability on a graphics processor

Dr. Multicast: Rx for Data Center Communication Scalability

Delay scheduling: a simple technique for achieving locality and fairness in cluster scheduling

A shared-subspace learning framework for multi-label classification

Tuning the capacity of search engines

A view of cloud computing

Clearing the clouds away from the true potential and obstacles posed by this computing capability.

Quantitative Analysis of Culture Using Millions of Digitized Books

We constructed a corpus of digitized texts containing about 4% of all books ever printed. Analysis of this corpus enables us to investigate cultural trends quantitatively. We survey the vast terrain of 'culturomics,' focusing on linguistic and cultural phenomena that were reflected in the English language between 1800 and 2000. We show how this approach can provide insights about fields as diverse as lexicography, the evolution of grammar, collective memory, the adoption of technology, the pursuit of fame, censorship, and historical epidemiology. Culturomics extends the boundaries of rigorous quantitative inquiry to a wide array of new phenomena spanning the social sciences and the humanities.

Exploratory community sensing in social networks

Inferring compositional style in the neo-plastic paintings of Piet Mondrian by machine learning

Case Theory: A Historical Overview

OverSoc: Social Profile Based Overlays

Interactive analysis and simulation of VANETs using MOWINE

We present a mobile wireless network evaluation tool (MOWINE) that can analyze mobility and wireless connectivity traces to quantify performance of different network protocols in a given application scenario. We model the highly dynamic connectivity of mobile networks using time-expanded graphs which allow us to study the best-case performance of network protocols. MOWINE also integrates with a network simulator that can evaluate the performance of recent VANET routing protocols with a user specified network load. This combination of network modeling and simulation enables users to gain deeper insights into the performance of routing protocols, for example, by distinguishing the limitations of a particular routing protocol from the fundamental limitations of the underlying network. Network engineers can use MOWINE to study and fine-tune performance of real-world network protocols in different scenarios before deploying the network, and hence engineer the mobile ad hoc network. We demonstrate the effectiveness of MOWINE by analyzing and understanding the performance of six different VANET routing protocols using real-world taxi cab traces collected in San Francisco.

Directing Crowd Simulations Using Navigation Fields

We present a novel approach to direct and control virtual crowds using navigation fields. Our method guides one or more agents toward desired goals based on guidance fields. The system allows the user to specify these fields by either sketching paths directly in the scene via an intuitive authoring interface or by importing motion flow fields extracted from crowd video footage. We propose a novel formulation to blend input guidance fields to create singularity-free, goal-directed navigation fields. Our method can be easily combined with the most current local collision avoidance methods and we use two such methods as examples to highlight the potential of our approach. We illustrate its performance on several simulation scenarios.

Virtualized Traffic: Reconstructing Traffic Flows from Discrete Spatiotemporal Data

We present a novel concept, Virtualized Traffic, to reconstruct and visualize continuous traffic flows from discrete spatiotemporal data provided by traffic sensors or generated artificially to enhance a sense of immersion in a dynamic virtual world. Given the positions of each car at two recorded locations on a highway and the corresponding time instances, our approach can reconstruct the traffic flows (i.e., the dynamic motions of multiple cars over time) between the two locations along the highway for immersive visualization of virtual cities or other environments. Our algorithm is applicable to high-density traffic on highways with an arbitrary number of lanes and takes into account the geometric, kinematic, and dynamic constraints on the cars. Our method reconstructs the car motion that automatically minimizes the number of lane changes, respects safety distance to other cars, and computes the acceleration necessary to obtain a smooth traffic flow subject to the given constraints. Furthermore, our framework can process a continuous stream of input data in real time, enabling the users to view virtualized traffic events in a virtual world as they occur. We demonstrate our reconstruction technique with both synthetic and real-world input.

On $L_{\infty}$ Properties of Multiresolution Scalar Quantizers

Modular Model Checking of Large Asynchronous Designs with Efficient Abstraction Refinement

Discovering Influence in Communication Networks Using Dynamic Graph Analysis

Towards a Formal Definition of a Computing Cloud

Cloud computing has been endorsed by the IT community as the new paradigm shift in the industry that charts the way forward. Unfortunately, the field is still on its path to rigor and robustness. This is epitomized by the numerous fuzzy articulations of “what is cloud computing”. This paper makes a first attempt at remedying this conundrum by providing a core technical specification of the model for cloud computing and demonstrating how current and future cloud deployments can use this to foster more productive technical discussion in future.

Superhuman performance of surgical tasks by robots using iterative learning from human-guided demonstrations

In the future, robotic surgical assistants may assist surgeons by performing specific subtasks such as retraction and suturing to reduce surgeon tedium and reduce the duration of some operations. We propose an apprenticeship learning approach that has potential to allow robotic surgical assistants to autonomously execute specific trajectories with superhuman performance in terms of speed and smoothness. In the first step, we record a set of trajectories using human-guided backdriven motions of the robot. These are then analyzed to extract a smooth reference trajectory, which we execute at gradually increasing speeds using a variant of iterative learning control. We evaluate this approach on two representative tasks using the Berkeley Surgical Robots: a figure eight trajectory and a two handed knot-tie, a tedious suturing sub-task required in many surgical procedures. Results suggest that the approach enables (i) rapid learning of trajectories, (ii) smoother trajectories than the human-guided trajectories, and (iii) trajectories that are 7 to 10 times faster than the best human-guided trajectories.

Do BitTorrent-Like VoD Systems Scale under Flash-Crowds?

Verifiable Encryption for P2P Block Exchange

Online Video Using BitTorrent and HTML5 Applied to Wikipedia

Improving Accuracy and Coverage in an Internet-Deployed Reputation Mechanism

Addressing the P2P Bootstrap Problem for Small Overlay Networks

Peer-to-Peer (P2P) overlays provide a framework for building distributed applications consisting of few to many resources with features including self-configuration, scalability, and resilience to node failures. Such systems have been successfully adopted in large-scale Internet services for content delivery networks, file sharing, and data storage. In small-scale systems, they can be useful to address privacy concerns as well as support for network applications that lack dedicated servers. The bootstrap problem, finding an existing peer in the overlay, remains a challenge to enabling these services for small-scale P2P systems. In large networks, the solution to the bootstrap problem has been the use of dedicated services, though creating and maintaining these systems requires expertise and resources, which constrain their usefulness and make them unappealing for small-scale systems. This paper surveys and summarizes requirements that allow peers potentially constrained by network connectivity to bootstrap small-scale overlays through the use of existing public overlays. In order to support bootstrapping, a public overlay must support the following requirements: a method for reflection in order to obtain publicly reachable addresses, so peers behind network address translators and firewalls can receive incoming connection requests; communication relaying to share public addresses and communicate when direct communication is not feasible; and rendezvous for discovering remote peers, when the overlay lacks stable membership. After presenting a survey of various public overlays, we identify two overlays that match the requirements: XMPP overlays, such as Google Talk and Live Journal Talk, and Brunet, a structured overlay based upon Symphony. We present qualitative experiences with prototypes that demonstrate the ability to bootstrap small-scale private structured overlays from public Brunet or XMPP infrastructures.

Multiparadigm Programming in Industry: A Discussion with Neal Ford and Brian Goetz

Guest Editors' Introduction: Multiparadigm Programming

Toward Predicting Collective Behavior via Social Dimension Extraction

A Chat Application in Lift

Algorithms for Advanced Battery-Management Systems

Multi-level Crypto Disk: Secondary Storage with Flexible Performance Versus Security Trade-offs

Peer Selection Strategies for Improved QoS in Heterogeneous BitTorrent-Like VoD Systems

Sepidar: Incentivized Market-Based P2P Live-Streaming on the Gradient Overlay Network

Hardware accelerated convolutional neural networks for synthetic vision systems

In this paper we present a scalable hardware architecture to implement large-scale convolutional neural networks and state-of-the-art multi-layered artificial vision systems. This system is fully digital and is a modular vision engine with the goal of performing real-time detection, recognition and segmentation of mega-pixel images. We present a performance comparison between a software, FPGA and ASIC implementation that shows a speed up in custom hardware implementations.

Convolutional networks and applications in vision

Intelligent tasks, such as visual perception, auditory perception, and language understanding require the construction of good internal representations of the world (or ”features”), which must be invariant to irrelevant variations of the input while, preserving relevant information. A major question for Machine Learning is how to learn such good features automatically. Convolutional Networks (ConvNets) are a biologically-inspired trainable architecture that can learn invariant features. Each stage in a ConvNets is composed of a filter bank, some non-linearities, and feature pooling layers. With multiple stages, a ConvNet can learn multi-level hierarchies of features. While ConvNets have been successfully deployed in many commercial applications from OCR to video surveillance, they require large amounts of labeled training samples. We describe new unsupervised learning algorithms, and new non-linear stages that allow ConvNets to be trained with very few labeled samples. Applications to visual object recognition and vision navigation for off-road mobile robots are described.