- talks - lists.cs.princeton.edu

Colloquium Speaker: Frank Pfenning Monday, September 26th
by Nicole E. Wagenblast 20 Sep '16

20 Sep '16

Colloquium Speaker Professor Frank Pfenning, Carnegie Mellon University Monday, September 26th, 12:30pm Computer Science 105 Design and Implementation of a Freshman Algorithms Course with Contracts We provide an overview of the introductory computer science curriculum at Carnegie Mellon University and then provide some detail about its perhaps most unusual component: a first-semester freshman-level course on imperative programming, algorithms, data structures, and computational thinking. A key aspect of this course is its use of C0, a small safe subset of C augmented with a language for expressing contracts that capture pre- and post-conditions as well as loop and data structure invariants. We provide some sample lecture material and report on our experience with this course over the last six years. Frank Pfenning studied Mathematics and Computer Science at the Technical University Darmstadt and then left for Carnegie Mellon University on a Fulbright scholarship where he obtained his Ph.D. in Mathematics in 1987 under the supervision of Professor Peter Andrews. He subsequently joined the Department of Computer Science at Carnegie Mellon University as research faculty where he became Professor in 2002 and served as Director of Graduate Programs from 2004 to 2008 and Associate Dean for Graduate Education from 2009 to 2010. He was appointed Head of the Computer Science Department in January 2013 and the Joseph F. Traub Professor of Computer Science in October 2015. He has spent time as visiting scientist at the Max-Planck-Institute for Computer Science in Saarbrücken, as Alexander-von-Humboldt fellow at the Technical University Darmstadt, and as visiting professor at École Polytechnique and INRIA-Futurs. He has advised 24 completed Ph.D. theses and won the Herbert A. Simon Award for Teaching Excellence in the School of Computer Science in 2002. He served as trustee, vice president, and president of CADE, Inc., the governing body of the International Conference on Automated Deduction, and on advisory boards for INRIA, the Max-Planck-Institute for Computer Science, and Seoul National University. He has chaired several conferences and program committees, including CADE and LICS, and has been a member of the editorial boards for Theoretical Computer Science, Journal of Automated Reasoning, and the Journal of Symbolic Computation. He was named Fellow of the ACM in 2015. His research interests include programming languages, logic and type theory, logical frameworks, automated deduction, and computer security. In his spare time he enjoys playing squash, running, hiking, cooking, and reading.

1 0

Allison Chaney will present her FPO, Computational Methods for Exploring Human Behavior", on Thursday, 9/8/2016 at 3:00pm in CS 402.
by Nicki Gotsis 01 Sep '16

01 Sep '16

Allison Chaney will present her FPO, Computational Methods for Exploring Human Behavior", on Thursday, 9/8/2016 at 3:00pm in CS 402. The members of her committee are David Blei (adviser), readers: David Blei and Barbara Engelhardt; Examiners: David Blei, Elad Hazan, and Brandon Stewart (Sociology). A copy of her thesis is available in Room 310. Everyone is invited to attend her talk. The abstract follows below: Researchers and analysts from many diverse fields are interested in unstructured observations of human behavior; this variety of data is constantly increasing in quantity. In this dissertation, we describe a suite of computational methods to assist investigators in interpreting, organizing, and exploring this data. We develop two Bayesian latent variable models for human-centered applications; specifically, we rely on additive Poisson models, which allow behavior to be associated with various sources of influence. Given observed data, we estimate the posterior distributions of these models with scalable variational inference algorithms. These models and inference algorithms are validated on real-world data. Developing statistical models and corresponding inference algorithms only addresses part of the needs of investigators. Non-technical researchers faced with analyzing large quantities of human behavior data are not able to use the results of inference algorithms without tools to translate estimated posterior distributions into accessible visualizations, browsers, or navigators. We present visualization based on an underlying statistical model as a first-class research problem, and provide principles to guide the construction of these systems. We demonstrate these principles with exploratory tools for two latent variable models. By considering the interplay between developing statistical models and tools for visualization, we are able to develop computational methods that provide for the full needs of investigators interested in exploring human behavior.

1 1

EE Seminar - August 24, 2016, 3:00 pm, E-Quad B205 - Dr. Eric Chang
by Nicole E. Wagenblast 17 Aug '16

17 Aug '16

DEPARTMENT OF ELECTRICAL ENGINEERING SEMINAR SERIES Speaker: Dr. Eric Chang Microsoft Research Asia Title: Age of AI: Artificial Intelligence Date: Wednesday, August 24, 2016 Time: 3:00 p.m. Room: B205 Engineering Quadrangle Host: Prof. S. Y. Kung Abstract: The building of a computer that can intelligently carry out tasks that require “human level intelligence” has been a goal for computer scientist since the 1950’s. Recently, tasks that were considered to exhibit human level intelligence such as understanding natural language, carry out conversational speech, playing master level chess, and interpreting and understanding images have all seen a great deal of improvements and many have seen real world applications. Nowadays it is rare not to see media coverage of AI on day-to-day basis. In this talk, I will provide an overview of the recent progress of artificial intelligence and highlight the gap between today’s artificial intelligence and human intelligence, predict the new future directions of AI and address several issues potentially raised by the further development of AI. Biography: Dr. Eric Chang joined Microsoft Research Asia (MSRA) in July, 1999 to work in the area of speech technologies. Eric is currently the Senior Director of Technology Strategy at MSR Asia, where his responsibilities include communications, IP portfolio management, and driving new research themes such as eHealth. Prior to joining Microsoft, Eric had worked in Nuance Communications, MIT Lincoln Laboratory, Toshiba, and General Electric Corporate Research and Development. Eric graduated from MIT with Ph.D., Master and Bachelor degrees, all in the field of electrical engineering and computer science. Eric’s work has been reported by Wall Street Journal, Technology Review, and other publications. Save

1 0

FANGFEI LIU (ELE) Time Change: TODAY 2:30PM, Room B327 Engineering Quadrangle
by Nicki Gotsis 17 Aug '16

17 Aug '16

Department of Electrical Engineering Final Public Oral Examination of Fangfei Liu Cache Side Channel Attacks and Secure Cache Architectures Wednesday, August 17, 2016 Room B327 Engineering Quadrangle 2:30PM All those interested are welcome to attend.

1 0

Ohad Fried's Pre FPO will take place on Thursday, July 14, 2016 (today) at 2pm in CS 402
by Nicki Gotsis 14 Jul '16

14 Jul '16

Ohad Fried's Pre FPO will take place on Thursday, July 14, 2016 (today) at 2pm in CS 402. The members of his committee are as follows: Adam Finkelstein (adviser)' Readers: Eli Shechtman (Adobe), Thomas Funkhouser; Non readers: Szymon Rusinkiewicz, David Dobkin. Everyone is invited to attend his talk. The talk title and abstract follow below: Title: "Photo Manipulation (The Easy Way)" Outline: With the growing number of images taken every second, and the growing size of personal photo collections, the need arises for more sophisticated yet easy to use photo manipulation techniques. This talk will present several projects that aim to “democratize” image manipulation: allowing sophisticated edits that are accessible to novices. I will start by presenting a novel way to arrange object collections, such as personal photo libraries. Collections are often presented visually in a grid because it is a compact representation that lends itself well for search and exploration. Most grid layouts are sorted using very basic criteria, such as date or filename. We present a method to arrange collections of objects respecting an arbitrary distance measure. Pairwise distances are preserved as much as possible, while still producing the specific target arrangement which may be a 2D grid, the surface of a sphere, a hierarchy, or any other shape. We show that our method can be used for infographics, collection exploration and data visualization. We present a fast algorithm that can work on large collections and quantitatively evaluate how well distances are preserved. Next, I will show a method to remove unwanted distracting element from images. We propose a new computer vision task — “distractor prediction”. Distractors are the regions of an image that draw attention away from the main subjects and reduce the overall image quality. Removing distractors — for example, using in-painting — can improve the composition of an image. We created two datasets of images with user annotations to identify the characteristics of distractors. We use these datasets to train an algorithm to predict distractor maps. Finally, we use our predictor to automatically enhance images. Lastly, I will show a method to modify the apparent relative pose and distance between camera and subject given a single portrait photo. Our approach fits a full perspective camera and a parametric 3D head model to the portrait, and then builds a 2D warp in the image plane to approximate the effect of a desired change in 3D. We show that this model is capable of correcting objectionable artifacts such as the large noses sometimes seen in “selfies,” or to deliberately bring a distant camera closer to the subject. This framework can also be used to re-pose the subject, as well as to create stereo pairs from an input portrait. We show convincing results on both an existing dataset as well as a new dataset we captured to validate our method.

1 0

Fisher Yu will present his Pre-FPO on Monday, July 18, 2016 at 1pm in CS 402
by Nicki Gotsis 13 Jul '16

13 Jul '16

Fisher Yu will present his Pre-FPO on Monday, July 18, 2016 at 1pm in CS 402. The members of his committee are: Thomas Funkhouser (adviser), Vladlen Koltun (reader;Intel), Szymon Rusinkiewicz (reader), Adam Finkelstein (non-reader), Kai Li (non-reader). Everyone is invited to attend his talk. The talk title and abstract follow below: Title: Pixel-Level Prediction of Depth and Semantics in Images Outline: Pixel-Level prediction is a common task in computer vision. Human eyes can assign each pixel 3D properties, such as depth and normal, and semantic meaning, such as object category and boundary. The pixel-level information can help algorithm process online content with finer granularity and help autonomous agents navigate in complicated environment. I will present my research in 3D geometry to predict pixel depth in previously impossible cases and convolutional networks that can improve image segmentation results substantially. I will also talk about my efforts to create large scale image dataset which may create new possibilities in computer vision research. I will first talk about pixel-level depth prediction from a set of images with very small motion. The images are taken from videos recording hand motion when he/she tries to hold the camera still. The scenario is challenging due to very small baseline. We analyze the SfM problem from geometry perspective and find an algorithm that can recover camera motion for most of the cases. Further, a multi-view stereo method robust to depth noise is developed to predict the depth at each pixel. We also show that the recovered depth can be useful for some computational photography applications. To predict the semantics of each pixel, or to do semantic image segmentation, we turn to convolutional networks, which is more powerful than the alternatives. A lot of works have studied how to use the network designed for image classification as basis to solve the segmentation problem. I argue that in order to design a better model for image segmentation, we have to change the basis to make it more suitable for segmentation. Toward this goal, I introduce dilated convolution as an additional building block for semantic segmentation and study its properties. We use dilated convolution to build a deeper network that has much wider receptive field than the original image classification network. Our model has got competitive performance on different datasets. Finally, I will introduce my on-going project, LSUN, which aims to make it possible to explore weakly annotated data to help solve the image segmentation problem. In LSUN project, we have collected orders of magnitude more labeled and unlabeled images than any previous datasets. To reach the scale, we design an iterative method to combine the advantages of classification models and human labeling. We also introduce additional statistical tests along the process so that the final labeled set has controlled precision. Up to now, we are able to collect around 70 million labeled images with about 90% labeling accuracy. Additional experiments show that LSUN dataset is poised to give us more insight in understanding images.

1 0

Sema Berkiten will present her FPO, "Capturing, Processing, and Synthesizing Surfaces with Details" on Thursday, 7/21/2016 at 2:00 pm in CS 402.
by Nicki Gotsis 13 Jul '16

13 Jul '16

Sema Berkiten will present her FPO, "Capturing, Processing, and Synthesizing Surfaces with Details" on Thursday, 7/21/2016 at 2:00 pm, CS 402. The members of her committee are: Szymon Rusinkiewicz (adviser), readers: Adam Finkelstein and Hao Li (University of Southern California); and nonreaders: Thomas Funkhouser and David Dobkin. A copy of her thesis is available in Room 310. Everyone is invited to attend her talk. The talk abstract follow below. Geometry acquisition and processing have become increasingly popular in computer graphics and vision, with demand for high-quality models driven by advances in 3D printing, realistic real-time renderings of 3D avatars in video games, digital libraries for historical objects etc. In this thesis, we focus on techniques to produce and process detailed geometry including acquisition of real world objects, processing and fusing the captured data, and synthesizing new surfaces from existing ones. First, we summarize a 2D acquisition technique called photometric stereo to capture high resolution surface details. As a validation step, we solve a misalignment problem for photometric datasets. After the dataset is validated, the surface normals can be computed using one of the photometric stereo algorithms. In order to decide which algorithm to use, we present a synthetic photometric benchmark to evaluate various algorithms for different scenarios. To produce a detailed surface in 3D, we propose an approach to combine a rough 3D geometry with detailed normal maps obtained from different views. We begin with unaligned 2D normal maps and a rough 3D geometry, and automatically align each normal map to the 3D geometry. Next, we map the normals onto the surface, correct and seamlessly blend them together. We then optimize the geometry to produce a high-quality 3D model. Next, we introduce a semi-automated system to convert photometric datasets into geometry-aware non-photorealistic illustrations of surface details that obey the common conventions of epigraphy (black-and-white archaeological drawings of inscriptions). This system is composed of rectification of the surface normals to correct camera perspective, segmentation of the inscriptions from the background, classification of the inscription based on carving technique, and stylization of the inscriptions in various styles. Finally, we present an algorithm for realistically transferring surface details (specifically, displacement maps) from existing high-quality 3D models to simple shapes that may be created with easy-to-learn modeling tools. Our key insight is to use metric learning to find a combination of geometric features that successfully predicts detail-map similarities on the source mesh, and use the learned feature combination to drive the detail transfer

1 0

Muneeb Ali is presenting his pre FPO on Thursday, June 30, 2016 at 3pm in CS 402
by Nicki Gotsis 28 Jun '16

28 Jun '16

Muneeb Ali is presenting his Pre FPO on Thursday, June 30, 2016 at 3pm in CS 402. The members of his committee are Andrea LaPaugh (advisor), Mike Freedman (reader), Arvind Narayanan (reader), Jen Rexford (non-reader), and Brian Kernighan (non-reader). Title and abstract follow below: Title: "Bootstrapping Trust in Distributed Systems with Blockchains" Abstract: Blockchains like Bitcoin and Ethereum have seen significant adoption in the past few years. Beyond their cryptocurrency uses, blockchains are being used to build new, decentralized versions of DNS and public-key infrastructure (PKI) that have no central points of trust. Blockchains can provide the basis for a general-purpose "trust layer" for Internet applications, but certain aspects of contemporary blockchains present technical limitations. Blockchains are hard to modify after initial deployment, they cannot store much data, they have slow writes (order of minutes), and the time to bootstrap new nodes increases linearly every day. Further, little production data or experience is available to guide design tradeoffs. In this thesis, we present the design of a new blockchain-based naming and storage system called Blockstack. Blockstack solves several limitations of contemporary blockchains and its design is informed by our experience of operating a large deployment of a decentralized PKI service built on top of the Namecoin blockchain. Our production deployment led to the discovery of a critical security problem on the Namecoin network along with highlighting network reliability and throughput challenges, while registering and updating over 33,000 entries. We present the design of (a) a migration framework for migrating from one blockchain to another under a failure of the underlying blockchain — our migration from Namecoin to Bitcoin used this framework and is the first crosschain migration of a production system running on blockchains, (b) a logically separate layer, virtualchain, which introduces novel new functionality to production blockchains without requiring any consensus-breaking changes from the underlying blockchain, and (c) a decentralized storage mechanism that is resistant to sybil-attacks. Blockstack is used as a general-purpose "trust layer" by a variety of applications/services, including a production PKI system for 57,000 users that has no single point of trust/failure and a large decentralized marketplace. Blockstack is released as open source software (http://github.com/blockstack).

1 0

Xin Jin will present his FPO "Dynamical Control of Software-Defined Networks" on Wednesday, 6/29/2016 at 2pm in CS 401.
by Nicki Gotsis 22 Jun '16

22 Jun '16

Xin Jin will present his FPO "Dynamical Control of Software-Defined Networks" on Wednesday, 6/29/2016 at 2pm in CS 401. The members of his committee are Jennifer Rexford (adviser), Readers: David Walker and Ratul Mahajan (Microsoft Research); Nonreaders: Jennifer Rexford, David Walker, Aarti Gupta, Ratul Mahajan (Microsoft Research), and Nick Feamster. A copy of his thesis is available in Room 310. Everyone is invited to attend his talk. The abstract follow below. Network management is critical to provide fast, reliable and secure network services. Software-defined networking (SDN) is a new network architecture to simplify network management by integrating network control to a centralized control platform. Network operators run various applications on the control platform to perform different management tasks, like routing, monitoring, load balancing and firewall. These applications have complex interactions with each other, making it difficult to deploy and reason about their behaviors. The frequent network events, such as traffic shifts, cyber attacks, and device failures, further exacerbates the problem. Each application needs to reconfigure the network, in order to react to the events. It is challenging to correctly and efficiently combine configuration changes from multiple applications, distribute these changes to a distributed collection of network devices, and coordinate changes across network devices in different layers. In this thesis, we present a new control architecture that can efficiently handle network events for multiple applications and across the network and optical layers. We identify and study the following three key components of the architecture. (i) CoVisor: A network hypervisor that can compose multiple applications and can effi- ciently merge configuration changes from these applications in the face of network events. To protect the network from malicious and buggy applications, CoVisor also provides topology virtualization and fine-grained access control to constrain what each application can see and do. (ii) Dionysus: A network update scheduler that can quickly and consistently distribute configuration changes to multiple switches. Dionysus uses a dependency graph to capture the dependencies between update operations, and dynamically schedules the operations based on runtime conditions. The approach both eliminates undesirable transient behaviors like loops, blackholes and congestion, and reduces the update time. (iii) Owan: A traffic management system that can jointly control the optical and network layers. Owan optimizes optical circuit setup, routing and rate allocation together, and dynamically adapts them to workload changes. The joint management significantly improves data transfers over the wide area network. We have built software controllers and hardware testbeds, and evaluated them with prototype experiments and large-scale simulations using network topologies and traffic traces from production networks.

1 0

Talk: James H. Graham, Today- 10am
by Nicole E. Wagenblast 22 Jun '16

22 Jun '16

Room: 402 Date: Wednesday, June 22 Time: 10am Host: Andrew Appel Trusted Software for Cybersecurity in Industrial Control Systems James H. Graham University of Louisville, Dept. of Electrical and Computer Engr. and True Secure SCADA, LLC This talk will discuss the use of seL4 and trusted software components in industrial control systems. The goal is to protect critical infrastructure installations against devastating cyber-attacks on the control systems which monitor, regulate and control these facilities (think for a moment about being without electrical power and water for a really long time!) This talk will include an overview of three successful attacks against supervisory control and data acquisition (SCADA) systems and a discussion of some approaches to mitigating these threats and their potential consequences through the use of trusted software components. Dr. James Graham’s research on ICS cybersecurity has been funded for ten years by the Dept. of Justice and the Dept. of Homeland Security and other national and state agencies. He worked as a product and process engineer for General Motors, and served as an engineering professor at Rensselaer Polytechnic Institute and at the University of Louisville for 34 years prior to retirement in 2014. He is co-founder and CEO of True Secure SCADA, LLC. _______________________________________________ talks mailing list talks(a)lists.cs.princeton.edu To edit subscription settings or remove yourself, use this link: https://lists.cs.princeton.edu/mailman/listinfo/talks _______________________________________________ talks mailing list talks(a)lists.cs.princeton.edu To edit subscription settings or remove yourself, use this link: https://lists.cs.princeton.edu/mailman/listinfo/talks

1 0