http://research.microsoft.com/apps/dp/vi/videos.aspx Watch the latest lectures from Microsoft Research © 2013 Microsoft Corporation. All rights reserved. en-US Sun, 09 Jun 2013 23:37:23 GMT Sun, 09 Jun 2013 23:37:23 GMT [Speaker: Tyler Furtwangler] Current Event: Taptitude - Designing for In App Purchase In this talk the developers of the hit mobile game Taptitude are going to share their keys to success. Learn how they modeled Taptitude for in-app-purchase, how they view free vs paid, and how to get users to hand over their hard earned money on your digital goods. Agenda: 6:30 PM – Presentation 7:30 PM – Ask the Experts, mingle, Q&A, etc. 8:00 PM – Event ends http://research.microsoft.com/apps/video/default.aspx?id=194281 Tyler Furtwangler Science and Technology Fri, 14 Jun 2013 01:30:00 GMT [Speakers: Shaileen Pokress and Andrew F. McKinney] Through an intuitive drag and drop interface that resembles snapping together LEGO blocks, MIT App Inventor makes it easy for even non-programmers to create apps for health, education, business, fun, social good and anything else that comes to mind. With almost 1 million people using App Inventor already, we're exploring ways to make it even more usable and accessible to people from all interests and backgrounds. App Inventor allows app developers to access all of the mobile device's features, providing an easy way to build fully functional apps that are personally relevant and useful. In this presentation, App Inventor Education Director Shaileen Pokress and Lead Software Architect Andrew McKinney will: Provide background information about the project Highlight interesting cases of App Inventor in action Go "under the hood" to explain how it works http://research.microsoft.com/apps/video/default.aspx?id=194283 Shaileen Pokress; Andrew F. McKinney Science and Technology Thu, 13 Jun 2013 17:30:00 GMT [Speaker: Martin Roetteler] One of the main challenges in quantum computing is to come up with new quantum algorithms, ideally for problems with business value, i.e., ultimately leading to commercialization. While there have been quite a few successes in quantum algorithms during the past two decades, arguably not much progress has been made by the research community toward a) fundamentally new techniques and b) really compelling business cases. In this talk, I will review some of my efforts to finding new applications for quantum computers, focusing on exponential speedups between classical and quantum computation. Quantum rejection sampling is an approach to quantum state generation that can be thought of as a quantum analog of the classical rejection sampling method (von Neumann, 1951). We assume that a black box is given that produces a coherent superposition of (possibly unknown) quantum states with some amplitudes. The problem is to prepare a coherent superposition of the same states, albeit with different target amplitudes. We exhibit a quantum algorithm for this problem and analyze its cost using semidefinite programming. As an application, we derive a quantum algorithm for the hidden shift problem for an arbitrary Boolean function. A bound on the runtime of this algorithm can be obtained by "water-filling" the Fourier spectrum. I will also give a brief synopsis of some of my other work in quantum computing, including the software development project TORQUE which provides tools for resource estimation. One of the main use cases of the TORQUE tool-suite is to to estimate the overheads and costs of quantum algorithm implementations on actual physical hardware. http://research.microsoft.com/apps/video/default.aspx?id=194286 Martin Roetteler Science and Technology Wed, 12 Jun 2013 17:30:00 GMT [Speaker: Maria Semple] Maria Semple’s new Seattle-based book, Where’d You Go, Bernadette, is a hilarious take on our fair city and all it has to offer — Craftsman houses, helicopter parents and, of course, the importance of tech in this town. None other than Microsoft plays a big part in this story of Bernadette, a MacArthur Genius Award-winning architect who loathes her adopted city, her husband, Elgin Branch, a brilliant gadget inventor at Microsoft, and their daughter Bee. Semple, a former TV writer talked to us about her book’s undeniably important background character – tech – and the role it plays in her book and Seattle. The everyday details on Microsoft are pretty amazing – how did you do the research to get all the information on what working there is like? It all appears very accurate and spot-on…It’s one of great compliments that I get now, and I’m just inundated with emails from Microsoft people who love the book and say that I got it exactly right, and they all love it, it’s very affectionate. Bernadette has a few quips about Microsoft, but it’s very affectionate for people inside Microsoft—people inside Microsoft love Microsoft. I thought that wouldn’t be true. I was actually surprised when I went there and everyone was into Microsoft. … I thought they’d be like, “Oh, I wish I worked at Apple.” I didn’t think they would be that happy, but I felt that was real, and I felt that I must represent that. http://research.microsoft.com/apps/video/default.aspx?id=194157 Maria Semple Science and Technology Tue, 11 Jun 2013 20:30:00 GMT [Speaker: Grigory Yaroslavtsev] Submodular functions capture the law of diminishing returns and can be viewed as a generalization of convexity to functions over the Boolean cube. Such functions arise in different areas, such as combinatorial optimization, machine learning and economics. In this talk we will give a brief overview of recent structural results about concise representations of submodular functions. Existence of small representations is useful for applications in learning such functions from examples and testing whether a given function is submodular with a small number of queries. For the class submodular functions taking values in discrete integral range of size R we show a structural result, giving concise representation for this class by formulas. The representation can be described as a maximum over a collection of threshold functions, each expressed by an R-DNF formula. This leads to efficient PAC-style learning algorithms for this class, as well as testing algorithms with running time independent of the size of the domain. Joint work with Sofya Raskhodnikova (SODA'13) and work in progress with Eric Blais, Krzysztof Onak and Rocco Servedio http://research.microsoft.com/apps/video/default.aspx?id=194081 Grigory Yaroslavtsev Science and Technology Tue, 11 Jun 2013 17:30:00 GMT [Speaker: Tim Hwang] The past decade has shown both the great strengths and great weaknesses of large scale, distributed crowd action. As we enter the next decade, how might we do better? This talk will take a wide-ranging, retrospective approach: examining the crowd in a variety of contexts from scaled funny cat production to large-scale political mobilization. It will also examine the intellectual history of the crowd, examining the arguments made by scholars in favor and against it. Based on this discussion, we will turn the possibility of cyborg-crowds – automated systems of various kinds that are built with shaping and optimizing the behavior of crowds in mind. In particular, we will examine experiments into the use of swarms of realistic automated identities online to shape the topology of social networks on a large scale. The ethics, economics, and future prospects of doing so will be examined alongside the engineering challenges. http://research.microsoft.com/apps/video/default.aspx?id=194072 Tim Hwang Science and Technology Mon, 10 Jun 2013 20:30:00 GMT [Speaker: Ian Kash] TBC http://research.microsoft.com/apps/video/default.aspx?id=194159 Ian Kash Science and Technology Fri, 07 Jun 2013 15:00:00 GMT [Speaker: Maria Polukarov] TBC http://research.microsoft.com/apps/video/default.aspx?id=194160 Maria Polukarov Science and Technology Fri, 07 Jun 2013 13:45:00 GMT [Speaker: Tobias Preis] TBC http://research.microsoft.com/apps/video/default.aspx?id=194162 Tobias Preis Science and Technology Fri, 07 Jun 2013 10:15:00 GMT [Speaker: Peter Key] TBC http://research.microsoft.com/apps/video/default.aspx?id=194164 Peter Key Science and Technology Fri, 07 Jun 2013 09:00:00 GMT [Speaker: Yagil Engel] We propose a new market design for display advertising contracts, based on posted prices. Our model and algorithmic framework address several major challenges: (i) the space of possible impression types is exponential in the number of attributes, which is typically large, therefore a complete price space cannot be maintained; (ii) advertisers are usually unable or reluctant to provide extensive demand (willingness-to-pay) functions, (iii) the levels of detail with which supply and demand are specified are often not identical. We propose a price space over compact set of market goods, each is a statement over a subset of the attributes. This implies a complex relationship between the way goods are represented in the market, and the goods on which players' utility functions are defined. We propose a market that supports this model, and prove that under reasonable assumptions on advertisers' utility equilibrium prices exist and can be computed in polynomial time. To facilitate elicitation, we show how utility functions can be extrapolated by observing advertiser's demand statements in response to current prices. Joint work with Moshe Tennenholtz. http://research.microsoft.com/apps/video/default.aspx?id=194166 Yagil Engel Science and Technology Fri, 07 Jun 2013 08:15:00 GMT [Speaker: Swarup Kumar Sahoo] Automatic Fault Localization by Filtering Likely Invariants An important step in repairing software failures discovered during testing or production runs is to diagnose the root cause(s) of the failure. Our work addresses automatic fault localization (i.e., identification of the program statements which are true root cause of a software failure) directly and extracts information useful in understanding why the root cause led to the failure. This work is based on three key ideas. First, likely program invariants can summarize program state in a compact way that can be compared efficiently between failing and successful runs, i.e., we can use likely invariants as a way to make delta debugging efficient. We select the program statements where likely invariants are violated during the failing run as an initial set of candidate root causes. Second, unlike previous work using invariants (for many different purposes), we create likely range invariants using automatically generated correct inputs that are close to the fault-triggering input to increase the likelihood of identifying the true root cause. Third, we reduce the number of false positives using some novel filtering techniques: dynamic backwards slicing, dependence filtering heuristic which removes an invariant from the candidate set if it is dependent on a previous failed invariant without any intervening passing invariant and finally filtering via multiple failing inputs that are also close to the failing input. Experimental results on reported software bugs in three widely-used large open-source applications Squid, Apache, and MySQL show that we are able to narrow down the number of candidate bug locations to between 5 and 17 program expressions, even in programs that are hundreds of thousands of lines long. Using manual inspection, we can quickly eliminate many of the remaining candidates because they are likely to violate invariants for incidental reasons (e.g., time-related functions, random-number generators, or input parsing etc.); remarkably, only 2-4 candidate locations remain after this manual filtering step without eliminating the true root cause. In a real tool, we can extract more information from the existing analyses like value of statements with failed invariants, control flow differences, passing and failing inputs to help programmers understand the root cause of the bug. In future, we plan to use concolic execution to develop a framework to construct good inputs fully automatically in an application-independent way and eliminate most false positives in a more robust and systematic manner. http://research.microsoft.com/apps/video/default.aspx?id=194074 Swarup Kumar Sahoo Science and Technology Thu, 06 Jun 2013 20:30:00 GMT [Speaker: Byung-Soo Choi] In this talk, I would like to discuss three research works. First, related to quantum applications, I would like to discuss the quantum partial search algorithms. Specifically, I will discuss how to generalize the single target block search case into the multiple target block search case. At the same time, I will show how to achieve sure success of the partial search. Second, related to quantum cryptanalysis, I would like to discuss how the quantum computer can be used for analyzing the Boolean functions, specially some secure properties. I will explain how Grover search algorithm can be used for analyzing the weight of Boolean functions where the weight is the ratio of solutions over the input. Its generalization for asymmetric case and multiple weights cases will be discussed. At the same time their optimality has been proved. Meanwhile for the actual secure primitives I will describe how the quantum computation can be used for resiliency checking problem. Third, related to quantum computer system, I would like to show an addition circuit on the 2D NTC architecture and their optimality. At the same time I will propose a way how to generated quantum LDPC code from any binary matrix. Also some current works on the layout and scheduling of logical tiles will be discussed. In each category, I will explain some future work. For example, quantum machine learning on the Boolean functions, quantum query/circuit complexities on the secure property checking, and the quantum error-correction code conversion methods will be touched. http://research.microsoft.com/apps/video/default.aspx?id=194288 Byung-Soo Choi Science and Technology Thu, 06 Jun 2013 17:30:00 GMT [Speaker: Rahul Savani] TBC http://research.microsoft.com/apps/video/default.aspx?id=194167 Rahul Savani Science and Technology Thu, 06 Jun 2013 14:30:00 GMT [Speaker: Dominic Berry] Simulation of physical quantum systems is potentially the most important application of quantum computers, and is Feynman's original motivation for proposing quantum computers. The first proposed techniques for simulation use Lie-Trotter product formulas, and are exponentially faster than classical algorithms in terms of the system dimension. However, they have worse scaling than classical algorithms in terms of other parameters: the system evolution time and the allowable error. I will present new techniques based on quantum walks and graduated compression, that provide improved scaling of the efficiency in terms of these parameters. http://research.microsoft.com/apps/video/default.aspx?id=194083 Dominic Berry Science and Technology Wed, 05 Jun 2013 20:30:00 GMT [Speaker: Dana Randall] Sampling algorithms using Markov chains arise in many areas of computation, engineering, and science. The idea is to perform a random walk among the elements in a large state space so that samples chosen from the stationary distribution are useful for the application. In order to get reliable results efficiently, we require the chain to be rapidly mixing, or quickly converging to equilibrium. Often there is a parameter of the system (typically related to temperature or fugacity) so that at low values many natural chains converge rapidly while at high values they converge slowly, requiring exponential time. This dichotomy is often related to phase transitions in the underlying models. In this talk we will explain this phenomenon, giving examples from the natural and social sciences, including magnetization, lattice gasses, colloids, and models of segregation. http://research.microsoft.com/apps/video/default.aspx?id=193883 Dana Randall Science and Technology Wed, 05 Jun 2013 20:00:00 GMT [Speakers: John Underkoffler and Kwindla Hultman Kramer] The future of computation will be characterized by real-time interaction with multiple screens and multiple devices. "Applications" will run as federated assemblies of processes executing across multiple CPUs. Graphics will render across multiple screens and displays. And architectures will enable harmonious use by more than one person at a time. Oblong Industries is a seven-year-old company founded with the goal of bringing about this future. Oblong's full-stack platform, called "g-speak", delivers multi-screen, multi-user UI components built on top of new networking paradigms and OS-level services. We'll briefly examine the technology's origins in complementary strands of research at the MIT Media Lab in the 1990s and its subsequent refinement through the Minority Report "prototyping experiment," and will then discuss some of the unique aspects of the architecture and the unconventional influences on its design. http://research.microsoft.com/apps/video/default.aspx?id=193909 John Underkoffler; Kwindla Hultman Kramer Science and Technology Wed, 05 Jun 2013 17:30:00 GMT [Speaker: Ramez Naam] We're beset by an array of natural resource and environmental challenges. They pose a tremendous risk to human prosperity, to world peace, and to the planet itself. Author Ramez Naam argues that the most valuable resource on earth is not oil, gold, water or land, but our capacity for expanding human knowledge. He charts a course to supercharge innovation - by changing the rules of our economy - that can lead the whole world to greater wealth and human well-being, even as we dodge looming resource crunches and environmental disasters and reduce our impact on the planet. http://research.microsoft.com/apps/video/default.aspx?id=193949 Ramez Naam Science and Technology Tue, 04 Jun 2013 20:30:00 GMT [Speaker: Arie Gurfinkel] Abstract interpretation (AI) is one of the most scalable automated program verification techniques. The scalability is achieved through aggressive abstraction in basic analysis steps (i.e., post, join and widening). This leads to loss of precision. As such, AI is plagued by false alarms. In this talk, I will present VINTA, an algorithm that enriches AI with Abstraction Refinement techniques from Software Model Checking to alleviate the false alarms. VINTA is an iterative algorithm that uses Craig interpolants to refine and guide AI away from false alarms. It is based on a novel refinement strategy that capitalizes on recent advances in SMT and interpolation-based Model Checking. On one hand, it can find concrete counterexamples to justify alarms produced by AI. On the other, it can strengthen invariants to exclude alarms that cannot be justified. The refinement process continues until either a safe inductive invariant is computed, a counterexample is found, or resources are exhausted. This strategy allows VINTA to recover precision lost in many AI steps. VINTA has been implemented as part of the UFO verification framework. It is a big contributor to the success of UFO in the 2nd International Software Verification Competition http://research.microsoft.com/apps/video/default.aspx?id=193912 Arie Gurfinkel Science and Technology Tue, 04 Jun 2013 20:30:00 GMT [Speaker: Julie Kientz] The use of interactive technologies to improve health and wellbeing has grown dramatically over the last two decades. However, there are many reasons why people still do not adopt different types of health technologies, including physical, mental, time, emotional, financial, and privacy demands. In my research, I have been seeking to understand and characterize these user burdens and design novel applications that can help to reduce them and improve access to healthcare. In this talk, I will first give an overview of studies seeking to understand the emotional, physical, and privacy burdens of interactive technologies. I will then describe the design and evaluation of three wellness applications my lab has developed in conjunction with health experts in which we have sought to reduce these burdens: 1) ShutEye, a mobile awareness display for promoting healthy sleep behaviors; 2) Lullaby, an at home capture and access system for monitoring the sleep environment; and 3) Baby Steps, an ecosystem of interactive tools for helping parents track developmental progress in young children. Finally, I will discuss future directions in helping to understand and reduce the user burden of health technologies. http://research.microsoft.com/apps/video/default.aspx?id=193915 Julie Kientz Science and Technology Tue, 04 Jun 2013 17:30:00 GMT