Skip to main content

Microsoft Research Audio

Microsoft

More than 1,100 brilliant scientists and engineers push the boundaries of computing in multiple research areas in 13 research labs around the world.



rss RSS

1,709
RESULTS


More right-solid

Show sorted alphabetically

More right-solid

Show sorted alphabetically

Unknown err: Trying to access array offset on value of type null [/home/tracey/petabox/www/common/Tiles.inc:501]
A B C D E F G H I J K L M N O P Q R S T U V W X Y Z
SHOW DETAILS
up-solid down-solid
eye
Title
Date Archived
Creator
Microsoft Research Audio
audio

eye 9

favorite 0

comment 0

This talk addresses the question of whether message passing provides an adequate programming model for multicore processors. It argues that programmers need more structure than what is provided by today's popular message passing libraries. Collective operations and design patterns offer some structure, but as the use of concurrency in programming increases, application programmers will have increasing difficulty identifying and combining these into complex operations. Some challenges, such as...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Jim Larus, Edward A. Lee
Microsoft Research Audio
audio

eye 0

favorite 0

comment 0

An important challenge in machine learning is how to efficiently learn from massive training data sets, especially with limited storage and computing capability. In this talk we introduce an efficient learning method called “compressed classification”, which aims to compress observations into a small number of pseudo-examples before classification. By analyzing the convergence rate of the …
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive
Multithreaded systems provide infrastructure capable of executing multiple software threads in parallel. Such systems range from the conventional multiprocessors to the recent simultaneous multithreading (SMT) and chip-multiprocessor (CMP) systems. Ideally, these systems should dynamically manage themselves such that applications always obtain the best performance. To achieve this goal, the systems must be able to self-monitor the runtime behavior of applications. In this talk, we focus on two...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Galen Hunt, Tong Li
While today's users work with and encounter a growing number and variety of computational devices (desktop PCs, laptops, tablets, PDAs, cellphones, etc.), continued adherence to the model of working with a single, personal computer has resulted in little support for coordinating activities across those devices. In fact, most devices are still completely unaware that a user might own other devices. In this talk I will describe my research group's work on supporting the shift from working with a...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Ken Hinckley, Jeff Pierce
Microsoft Research Audio
audio

eye 0

favorite 0

comment 0

The use of a virtual machine language, instead of machine code, is by now a well-established and successful technique for porting programs across different hardware platforms, without incurring into the difficulties and draw-backs of software distribution, when done at source-language level. In addition, interoperability between different source languages is made possible by their translation into …
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive
Microsoft Research Audio
audio

eye 3

favorite 0

comment 0

The lines between mobile devices - such as PDAs, mobile phones and cameras - have blurred, and are now almost gone. Similarly, the lines between the resulting technology and we, as human beings, are blurring. The challenge for the field of Computer Science, and in particular Human-Computer Interaction, is to ensure that the technology seamlessly fades into the fabric of our lives, rather than requiring ever more of our attention.
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Amy Karlson, Daniel Ashbrook
Microsoft Research Audio
audio

eye 1

favorite 0

comment 0

Research into Example Based Speech Recognition got revived over the past decade. Example based recognition is appealing for a number of reasons: there is considerable evidence from the linguistic literature that humans store actual traces of at least some sentences or phrases. Moreover, after 40 years of refining the HMM framework, we are still stuck …
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive
Microsoft Research Audio
by Microsoft Research
audio

eye 5

favorite 0

comment 0

Most modern cryptography, and 'public-key' crypto in particular, is based on mathematical problems that are conjectured to be infeasible (e.g., factoring large integers). Unfortunately, standard public-key techniques are often too inefficient to be employed in many environments; moreover, all commonly used schemes can in principle be broken by quantum computers. This talk will review my recent work on developing new mathematical foundations for cryptography, using geometric objects called...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Kristin Lauter, Chris Peikert
Microsoft Research Audio
by Microsoft Research
audio

eye 7

favorite 0

comment 0

Several academic and industry studies have documented substantial improvements in code design and defect rate when programs are written by pairs of programmers (two people working shoulder-to-shoulder at a single computer) versus individual programmers working independently (solo). Collateral benefits are reported to include improved morale and project knowledge shared efficiently across the development team in a manner that improves productivity in subsequent development cycle. In our...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Jim Larus, John Spencer, Larry Leifer,...
Without well-provisioned dedicated servers, modern fast-paced action games limit the number of players who can interact simultaneously to 16-32. This is because interacting players must frequently exchange state updates, and high player counts would exceed the bandwidth available to participating machines. This talk will describe Donnybrook, a system that enables epic-scale battles without dedicated server resources, even in a fast-paced game with tight latency bounds. It achieves this...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Jay Lorch, Jeffrey Pang
The progress of any creative discipline changes significantly with the quality of the tools available. As the diversity of user interfaces multiplies in the shift away from personal desktop computing, yesterday's tools and concepts are insufficient to serve the designers of tomorrow's interfaces. My research in human-computer interaction focuses on the earliest stages in UI creation - activities that take a novel idea and transform it into a concrete, interactive artifact that can be...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Andy Wilson, Bjoern Hartmann
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Jim Gemmell, Dan Ellis
Microsoft Research Audio
by Microsoft Research
audio

eye 6

favorite 0

comment 0

Unexpectedly, JavaScript has become one of the world's most popular programming language while still remaining one of the world's most misunderstood. What does the future hold for this much maligned and deeply flawed language? What can it teach us about the process of standards evolution? Can JavaScript ever become respected? Will JavaScript ultimately save the world? ©2008 Microsoft Corporation. All rights reserved.
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Wolfram Schulte, Douglas Crockford
Microsoft Research Audio
audio

eye 3

favorite 0

comment 0

XQuery is the upcoming standard language for querying XML data. The dynamic semantics of XQuery is defined in such a way that the evaluation of an XQuery expression may 'crash' (raise a run-time error). When an XQuery never crashes we say that it is well-defined. Automatic checking for well-definedness is, of course, undecidable. Nevertheless, for certain interesting fragments of XQuery that are similar in expressiveness to standard relational query languages, well-definedness can become...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Yuri Gurevich, Jan Van den Bussche
Microsoft Research Audio
audio

eye 6

favorite 0

comment 0

Model­-based software and system design is based on the end-to-end use of formal, composable and manipulable models in the product life-cycle. Model Integrated Computing (MIC) developed at the Institute for Software Integrated Systems (ISIS) at Vanderbilt University is part of this new direction together with other well known approaches and initiatives, such as Microsoft’s Software Factories, OMG’s Model Driven Architecture (MDA), or IBM’s Eclipse Modeling Framework (EMF). An emerging...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Yuri Gurevich, Janos Sztipanovits
Microsoft Research Audio
audio

eye 3

favorite 0

comment 0

In the past, the main methods for solving problems in Multiview Vision Geometry have been iterative techniques, which may suffer from falling into local minima, and trouble with convergence. Recent research has turned to finding guaranteed globally optimal solutions to such problems. Techniques include quasi-convex optimization, Second Order Cone Programming, Branch-and-bound techniques and fractional programming, solving many of the common vision geometry problems. In this talk, we address...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Sing Bing Kang, Richard Hartley
To understand Russian foreign policy goals after September 11 requires us to take much more seriously a factor generally downplayed in previous analyses: namely, debates among Russian elites concerning what it means for Russia to be a 'great power' in contemporary world politics. From this perspective, Putin’s emphasis on 'rebuilding the state' in order to restore Russia to its rightful global position has not only domestic, but also foreign policy implications. If ways can be found to...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Tamara Pesik, Steve Hanson
Microsoft Research Audio
audio

eye 4

favorite 0

comment 0

At the RAD Lab we are prototyping forward-looking datacenter software architectures using a three-pillar approach. The first pillar is exploiting application frameworks and languages optimized for high programmer productivity such as Ruby on Rails. Second is the deployment of machine learning to identify performance and scalability bottlenecks, create dynamic models for predicting performance, and mining runtime telemetry as well as console logs to identify operational problems; a framework we...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Jim Larus, Armando Fox
Retrieving information from online resources is an increasingly prevalent task, supporting many work-related and personal activities. Yet, despite the success of modern search engines, finding specific information can be difficult and time-consuming. When users who are experienced in a particular domain search for information, they often know what types of information can be found in particular types of documents. We are investigating a model, called semantic components, that seeks to leverage...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Ryen White, Susan Price
A piece of software is considered interactive if it consumes input and produces output throughout its execution, in contrast to non-interactive software which takes its input at program initialization and produces its output at program termination. Interactive software includes network servers, user interface applications, and computer games, and makes up the majority of user-facing software. The most common approach to implementing interactive components within modern languages is an...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Jim Larus, Brian Chin
In this work, we expose an unorthodox adversarial attack that exploits the transients of a system's adaptive behavior, as opposed to its limited steady-state capacity. We show that a well orchestrated attack could introduce significant inefficiencies that could potentially deprive a network element from much of its capacity, or significantly reduce its service quality, while evading detection by consuming an unsuspicious, small fraction of that element's hijacked capacity. This type of attack...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Solom Heddaya and Phil Fawcett, Mina...
Microsoft Research Audio
audio

eye 3

favorite 0

comment 0

An elliptic divisibility sequence is an integer recurrence sequence related to the multiples of a rational point on an elliptic curve. We generalise this definition to higher dimensions: an elliptic net is a map from Z n to Z encoding information about the Mordell-Weil group. This gives new methods of computation on elliptic curves, including a new algorithm to compute the Tate pairing for pairing-based cryptography. ©2007 Microsoft Corporation. All rights reserved.
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Kristin Lauter/Jennifer Chayes, Katherine...
Most combinatorial optimization problems of interest are NP-hard to solve exactly. To cope with this intractability, one settles for approximation algorithms with provable guarantee on the quality of approximation. Despite great success in designing approximation algorithms, underlying a vast majority of the work is the technique of linear programming, or more generally semi-definite programming. This poses the natural question: How far can we push the technique of semi-definite programming?...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Yuval Peres, Prasad Nagaraj Raghavendra
Microsoft Research Audio
audio

eye 3

favorite 0

comment 0

In this talk I will survey the latest development in dynamic algorithms for fundamental graph problems, such as shortest paths, reachability and connectivity. In the last five years there were many STOC/FOCS/SODA publications on dynamic graphs. I will overview the most important results in the area and will focus on two results regarding shortest paths which are part of my PhD thesis. ©2006 Microsoft Corporation. All rights reserved.
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Jennifer Chayes / Christian Borgs, Liam...
Microsoft Research Audio
audio

eye 2

favorite 0

comment 0

Driven by the hardware shift to multi-core architectures, concurrency is being brought into mainstream software development. Unfortunately, concurrent programs are prone to concurrency bugs, because of the inherent complexity of concurrency and the sequential thinking habits of programmers. Concurrency bugs' non- deterministic property also brings a lot of trouble to developers. Improving the reliability of concurrent programs is a critical and urgent task. This talk presents recent work on...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Tom Ball, Shan Lu
Microsoft Research Audio
audio

eye 0

favorite 0

comment 0

Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Liz Lawley, Danah Boyd, Clay Shirky
Modern Virtual Machines (VMs) generate rich runtime information that can often be used to perform self-optimization such as adaptive compilation. However, such valuable information has rarely been used by the underlying operating systems to create more optimized execution environments for applications running on these VMs. In this talk, I will report early results from our research to design OS schedulers that try to make favorable scheduling decisions to improve the overall performance of...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Ben Zorn, Witty Srisa-an
How much can randomness help computation? Motivated by this general question and by volume computation, one of the few instances where randomness probably helps, we analyze a notion of dispersion and connect it to asymptotic convex geometry. We obtain a nearly quadratic lower bound on the complexity of randomized volume algorithms for convex bodies in R n (the current best algorithm has complexity roughly n 4 and is conjectured to be n 3 ). Our main tools, dispersion of random determinants and...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Laszlo Lovasz, Santosh Vempala
Microsoft Research Audio
audio

eye 4

favorite 0

comment 0

Many Artificial Intelligence (AI) tasks, such as natural language processing, commonsense reasoning and vision, could be naturally modeled by a language and associated inference engine using both relational (first-order) predicates and probabilistic information. While logic has been the basis for much AI development and is a powerful framework for using relational predicates, its lack of representation for probabilistic knowledge severely limits its application to many tasks. Graphical models...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Eric Horvitz, Rodrigo de Salvo Braz
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Liz Lawley, Clay Shirky
With microprocessors hitting the power wall, alternative architectures are receiving substantial attention. Two factors make FPGAs promising, particularly for high performance computing applications: (i) the potential for thousand-fold parallelism, and (ii) the embedding of control into logic, which enables payload to be delivered every cycle. In this talk we will describe work in building FPGA-based accelerators for computational biology and bioinformatics. We find that, while there is...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Alessandro Forin, Yongfeng Gu
We present a technique which improves the Kneser- Ney smoothing algorithm on small data sets for bigrams and we develop a numerical algorithm which computes the parameters for the heuristic formula with a correction. We give motivation for the formula with correction on a simple example. Using the same example we show the possible difficulties one may run into with the numerical algorithm. Applying the algorithm to test data we show how the new formula improves the results on cross-entropy....
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Ye-Yi Wang, Peter Taraba
Microsoft Research Audio
audio

eye 2

favorite 0

comment 0

Modeling and estimating software reliability during testing is useful in quantifying the quality and dependability of the developed software systems. However, such measurements applied so late in the development process leave done little to improve the quality of the software system in a cost-effective way. Reliability, an important dependability attribute, is defined as the probability that the system performs its intended functionality under specified design limits. We argue that in order to...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Wolfram Schulte, Roshanak Roshandel
Microsoft Research Audio
audio

eye 1

favorite 0

comment 0

We present a library of generic software components for parallel and distributed computations on graphs, based on the Boost Graph Library (BGL). The BGL consists of a rich set of generic graph algorithms and supporting data structures, but it was not originally designed with parallelism in mind. In this talk, we revisit the abstractions comprising the BGL in the context of distributed-memory parallelism, lifting away the implicit requirements of sequential execution and a single shared address...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Kumar Chellapilla, Doug Gregor
Microsoft Research Audio
by Microsoft Research
audio

eye 2

favorite 0

comment 0

I will consider the design of efficient and profit-maximizing Bayesian incentive-compatible mechanisms for general dynamic environments with private information. In the environment, agents observe a sequence of private signals over a number of periods. In each period, the agents report their private signals and make public (contractible) and private decisions based on the reports. The probability distribution over future signals may depend on both past signals and past decisions. First I...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Jennifer Chayes, Ilya Segal
Microsoft Research Audio
audio

eye 5

favorite 0

comment 0

The SenseCam is a personal, wearable camera developed by Microsoft Research in Cambridge, UK, and used as a lifelogging device in projects like MyLifeBits. Its use in applications like MLB and and the memory aid research for Alzheimer suffers at Addenbrooks hospital in the UK, is based on wearing the SenseCam for lifelogging of 'events' during your day, and generating a fast-forward movie of the event as the memory recall interface. At Dublin City University we have been working with SenseCams...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Jim Gemmell, Alan Smeaton
Despite intensive research efforts directed toward improvement of nucleotide dye labelling technology and modification of the polymerase used for automated DNA sequencing, the peak heights seen in sequence electropherogram data still vary significantly between positions within a sequence. This is largely because polymerase enzymes incorporate dye labelled di-deoxynucleotides into the extending DNA chain at variable rates during cycle sequencing. In consequence, heterozygous positions rarely...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, David Heckerman & Nebojsa Jojic,...
Inferring regulatory mechanisms based on in silico analysis of regulatory elements has been the target of much research efforts in recent years. Specific aims include identifying combinatorial interactions of either known or novel transcription factors; identifying the cellular conditions in which these combinations are activated and which genes they control; pinpointing the active binding sites of each transcription factor; and finally predicting the expression profile of a gene given its...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, David Heckerman and Nebojsa Jojic, Yoseph...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Mike Wellings
For many predictive modeling tasks, acquiring supervised training data for building accurate classifiers is often difficult or expensive. Training data may either be limited, or often additional data may be acquired, but there is a cost associated with the acquisition. We study the problem of learning with reduced supervision in three setting. First, in the pure supervised learning setting, where we try to maximize the utility of small datasets. Second, in a traditional active learning setting,...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Eric Brill, Prem Melville
Microsoft Research Audio
audio

eye 0

favorite 0

comment 0

Myerson has argued that the Nash equilibrium lies at the foundations of modern economic thought, and yet the dark computational side of the concept keeps getting gloomier. We show that playing games under considerations of risk, disambiguating games via equilibrium selection a`-la Harsanyi-Selten, and computing equilibria by the homotopy method, are all rife with very …
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive
It’s the trap that ensnares virtually every business. We focus on process: “how” we’re doing the job and we forget the bigger issue: “what” we’re doing and “why” we’re doing it. We can rise above the “hows” to expose what does and doesn’t need attention in your organization, and to identify the activities most central to success. By looking at powerful case studies—from ING Direct to Amazon.com to Procter and Gamble and more-we can see how these diverse companies...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Kirsten Wiley, Ric Merrifield
Microsoft Research Audio
audio

eye 10

favorite 0

comment 0

We consider the asymmetric traveling salesman problem for costs satisfying the triangle inequality. We derive a randomized algorithm which delivers a solution within a factor O(log n/ log log n) of the optimum with high probability. Also we give the first constant factor approximation algorithm for metrics that are shortest path distances in a weighted directed graph when the underlying undirected graph has a bounded orientable genus. In this talk I will try to describe the main ideas of these...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Kamal Jain, Shayan Oveis Gharan
The human object recognition system has about a trillion synapses and computer vision systems will probably need to learn a similar number of parameters in order to be competitive. This makes it necessary to learn useful feature detectors from unlabeled examples (as the cortex does). I will describe how this can be done and will illustrate the approach with two multi-layer neural networks that perform object recognition on two very different databases. The NORB database has stereo, gray-level...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Li Deng and Alex Acero, Geoff Hinton
Much recent research has been devoted to learning algorithms for deep architectures such as Deep Belief Networks and stacks of auto-encoder variants, with impressive results obtained in several areas, mostly on vision and language data sets. The best results obtained on supervised learning tasks involve an unsupervised learning component, usually in an unsupervised pre-training phase. …
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive
Microsoft Research Audio
by Microsoft Research
audio

eye 8

favorite 0

comment 0

Nick Baker is General Manager for Xbox Architectural Design. After graduating from Imperial College London in 1990, he found his way to Apple and worked on the team that tried to create a specialized video card. He then went to 3DO where he worked on their high-end gaming system, which unfortunately failed in the market.
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Robert Hess, Nick Baker
In his talk, Professor Leavitt will focus on the philosophical, mathematical, and psychological underpinnings of Turing's landmark work, in particular Turing's proof of the insolubility of Hilbert's 'decision problem' (in the course of which he introduced the concept of the Turing machine) and the provocative essay on artificial intelligence in which Turing asked the famous question, 'Can a machine think?' Professor Leavitt will also discuss Turing's experiences as a homosexual man in pre- and...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Tamara Pesik, David Leavitt
Microsoft Research Audio
audio

eye 19

favorite 0

comment 0

How many days were there in 1752? Is there a year 0? When was there a February 30th? Why did an aluminum company lose $1,000,000 on New Year's Eve? Who thinks 1900 was a leap year? Why is Passover late? Why does Ramadan end early? Does the Chinese calendar follow a 19 year cycle? Why do the Hindus skip days and months? Who would pay for a Martian calendar? This talk will answer these ten questions and explain the underpinnings of the major types of calendars: solar (Gregorian, Julian, Persian,...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Yuri Gurevich, Nachum Dershowitz
Microsoft Research Audio
audio

eye 0

favorite 0

comment 0

Ever-increasing amounts of text are produced by end-users or collected from multiple sources. Since such data typically has many errors and lacks standardized representation, similarity query processing has recently drawn significant interests; it has a wide range of applications including query refinement for web search, near duplicate document detection and elimination. In this talk, we …
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive
Microsoft Research Audio
audio

eye 1

favorite 0

comment 0

3:30 – 4:00 Peter Binfield 4:00-5:00 John Wilbanks Keynote
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive
We introduce the formal foundations of a set-theoretic data model that can model data at both the logical and physical level. To demonstrate its practical value, we show how to use it to dynamically restructure data based on query requirements. Over time, most queries can be answered by retrieving from disk at most a small superset of the data they actually need, thereby yielding higher performance than conventional methods in today's database systems. The formal foundation defines operations...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Phil Bernstein, David Childs
In this talk, I will review use-inspired research being conducted in the IMPACT lab at Arizona State University (http://impact.asu.edu) focused on novel secure, safe and energy-efficient (thermal-aware) adaptive solutions for cyber-physical applications of embedded and networked wireless sensors and actuators in various domains including pervasive healthcare, smart cargo containers, precision agriculture, and sustainable datacenter computing. Exemplar solutions include physiological value based...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Johannes Helander, Sandeep Gupta
Since most interesting cryptographic tasks are impossible to achieve with absolute, information-theoretic security, modern cryptography aims to design cryptographic primitives (i.e., algorithms/protocols) that are computationally infeasible to break. Proving lower bounds of the type needed, however, seems beyond the reach of current techniques in complexity theory. Thus, research in the Foundations of Cryptography has aimed to design primitives based on complexity assumptions that are as weak...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Jennifer Chayes, Iftach Haitner
Structured classification deals with a family of problems where a response variable possessing meaningful internal structure has to be predicted from a set of input variables. This includes prediction problems involving strings, trees, or graphs as well as collective classification problems, i.e. problems where multiple correlated outputs have to be jointly predicted. We propose a general framework that combines the effectiveness of discriminative methods such as Gaussian processes and Support...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Joshua Goodman, Thomas Hofmann
Microsoft Research Audio
audio

eye 0

favorite 0

comment 0

Today, software is ubiquitous—it is deployed on virtually all electronic devices, small and large, including those that are life- and safety-critical. The need for robust, certifiably correct software requires us to develop the theory and tools for mechanically reasoning about, and also automatically generating, programs. In this talk, I will present the theory and practice …
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive
Microsoft Research Audio
audio

eye 2

favorite 0

comment 0

As networked systems grow in scale and complexity, system introspection becomes an increasingly important and challenging problem. Introspection is the ability to characterize system behavior, from identifying normal conditions to detecting any unexpected or undesirable events---attacks, configuration mistakes, security vulnerabilities, overload, or memory leaks due to buggy applications---before serious harm is done. However, to provide system introspection, monitoring services face two...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Jim Larus, Navendu Jain
Microsoft Research Audio
by Microsoft Research
audio

eye 0

favorite 0

comment 0

Using Machine Learning to Characterize, Predict and Optimize System Behavior - Presented by Archana Ganapathi Abstract: Systems have become very complex and it is often difficult to maintain an accurate model of various system components and their interactions. We use Statistical Machine Learning algorithms to automatically extract relationships between system workload and behavioral metrics. Using Kernel Canonical Correlation Analysis (KCCA) to extract correlations between system workload and...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Jim Larus, Kristal Sauer, Archana...
Microsoft Research Audio
audio

eye 0

favorite 0

comment 0

Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Liz Lawley, Group discussion
A popular trend in computer vision, graphics, and machine learning is to replace sophisticated statistical models with simpler generic ones, and to compensate for the missing domain knowledge with huge datasets. These huge datasets in turn require us to solve huge numerical optimization problems that tax popular off-the-shelf implementations of popular algorithms. I describe a randomized way to solve these large scale optimization problems quickly, in a few lines of code, and with provably good...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, John Platt, Ali Rahimi
Ad hoc networks are formed as collections of nodes that communicate over wireless channels. The dynamic and distributed nature of such networks means that randomization is often used to improve efficiency and to achieve symmetric solutions. In this talk, we demonstrate the applicability of an automated formal verification technique called probabilistic model checking to the analysis of ad hoc network protocols. A probabilistic model checker calculates the probability of a given temporal logic...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Jakob Rehof, Marta Kwiatkowska
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Craig Knoblock, Horst Bischof
The first half of this talk will discuss scalable methods for performing particular object search via visual queries in very large datasets of unordered images. We show how the combination of large vocabularies, fast spatial verification, query expansion and soft assignment can dramatically boost the accuracy of object retrieval to give a fast and reliable system. The second half will explore object mining in image datasets, where the aim is to automatically discover and group images containing...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Rick Szeliski, James Philbin
When we talk about analyzing vast amounts of consumer-generated media in order to find the 'gold' in specific topics, issues, trends, opinions and sentiment, we're talking about some powerful technologies that do the heavy lifting. Nielsen BuzzMetrics' content mining capabilities are rooted in machine-learning and natural language processing technologies that mine unstructured data—vast amounts of raw text—to discover the intelligence it contains. These technologies are able to identify key...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Michael Gamon, Navot Akiva
Mesh networks were proposed as a promising low cost, easy to deploy network architecture for wired and wireless networks. Mesh overlays provide the ideal setting to react quickly to fluctuating performance, use application-specific metrics when selecting paths, and support concurrent data flows over multiple paths. In addition, in the context of wireless communication, mesh networks offer increased coverage and larger bandwidth, resulting in higher quality of service and information...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Jitu Padhye, Cristina Nita-Rotaru
Although many audio signal processing techniques, developed for a wide range of applications such as denoising, source separation or time/pitch-scale modification, operate in the time-frequency power or magnitude domain, discarding the phase information is not without raising important issues. First, if resynthesis of a time-domain signal is necessary, phase needs to be estimated in such a way that it is coherent with the magnitude. Second, additivity of signals is not true anymore, as the...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Jasha Droppo, Jonathan Le Roux
Since the inception of electronic environments, researchers have been interested in how to provide better support for the tasks people perform in these environments. While several studies have examined behaviours such as revisitation and navigation on the Web, there has been little study of these behaviours in the context of task. There is, therefore, still a lack of understanding of the types of tasks people engage in on the Web and how people interact with their web browsers to complete these...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Kori Quinn, Melanie Kellar
Microsoft Research Audio
audio

eye 1

favorite 0

comment 0

Early design decisions in software projects profoundly affect both the properties and the costs of the eventual implementation. It is much easier and cheaper to change these decisions during design than after implementation has yielded running code. Improvements in our ability to predict properties of an implementation without actually inspecting the code would enable software designers to better understand the consequences of early decisions and would facilitate comparison of design...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, John Spencer, Wolfram Schulte, Mary Shaw
Microsoft Research Audio
audio

eye 1

favorite 0

comment 0

Traditional Operating System design has been primarily motivated by the need of optimizing system-wide efficiency factors defined in terms of average throughput or utilization. However, with the increase in processor speed and other resources, it is possible to provide better guaranty and predictability in the performance of individual applications, as perceived at the user level, without incurring significant loss in system-wide efficiency. A QoS-sensitive application (e.g., a multimedia...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Galen Hunt, Sourav Ghosh
Microsoft Research Audio
audio

eye 16

favorite 0

comment 0

For nearly a decade, John Wood rode the '90's tech boom to a leadership position at Microsoft where he 'had it all': the big paycheck, the corporate apartment and proximity to software giant Bill Gates. All of this changed with an eye opening trip to Nepal that inspired John to make a stunning decision: He would give up his career in order to build desperately needed schools across the developing world. With LEAVING MICROSOFT TO CHANGE THE WORLD Wood tells the remarkable story of his decision...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Kirsten Wiley and Kim Ricketts, John Wood
This presentation will focus on successful strategies for recruiting and retaining an outstanding and diverse engineering workforce. Examples include best practices for proactive search processes that yield a strong and diverse applicant pool. In addition, specific examples are given regarding how to conduct equitable and effective interviews. Of course, recruiting a talented and diverse team is only the first step in creating a diverse workplace. The real challenge is retaining that talent....
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Jennifer Chayes, Denice Denton
Microsoft Research Audio
audio

eye 1

favorite 0

comment 0

The SwitchWare project proposed to accelerate network evolution by Introducing programmable nodes into the network, enabling innovations to be introduced by network operators or even network users. An approach balancing flexibility, extensibility, security and performance was pursued using OCaml, resulting in the ALIEN architecture, which permitted construction of an incrementally deployed network element, the Active Bridge. ALIEN also enabled the construction of domain-specific languages such...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Jim Larus, Jonathan Smith
Microsoft Research Audio
audio

eye 2

favorite 0

comment 0

Texture synthesis has been widely recognized as an important research topic. In this talk, I present a series of graphics or vision related projects that focus on texture analysis and synthesis. These projects cover 2D texture synthesis, 3D texture reconstruction, tensor-based BTF compression, static surface texture synthesis and dynamic surface flow simulation. Specifically, in 2D texture synthesis, we have introduced a new technique called feature-based synthesis which can produce superior...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Brian Guenter, Yizhou Yu
Microsoft Research Audio
audio

eye 1

favorite 0

comment 0

Traditional Coding Theory, inspired by Shannon's motivations in communication problems, has been very successful in proposing meaningful abstractions (e.g Hamming distance, code rate, decoding complexity) and perfecting code construction for those abstractions. Data-storage applications, however, exhibit unique behaviors and constraints that are often addressed by either inefficient adaptations of known codes, or by ad-hoc solutions. In this lecture, examples of more methodical treatments of...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Phil Chou, Yuval Cassuto
Microsoft Research Audio
audio

eye 2

favorite 0

comment 0

Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Liz Lawley, Danah Boyd
Microsoft Research Audio
by Microsoft Research
audio

eye 2

favorite 0

comment 0

The goal of code obfuscation is to make a program completely 'unintelligible' while preserving its functionality. Obfuscation has been used for years in attempts to prevent reverse engineering, e.g., in copy protection and licensing schemes. Recently, spammers have utilized it to conceal code that spawns pop-ups. Finally, obfuscation is a cryptographer's dream: nearly any cryptographic task could be achieved *securely* by writing a simple program and then obfuscating it (if possible!). Barak et...
Topics: Microsoft Research, Microsoft Research Audio MP3 Archive, Jennifer Chayes / Christian Borgs, Yael...