- talks - lists.cs.princeton.edu

Colloquium Speaker Thomas Wenisch, Today, Dec 8, 4:30pm
by Nicole E. Wagenblast 08 Dec '14

08 Dec '14

Colloquium Speaker Thomas Wenisch, University of Michigan Monday, December 8, 4:30pm Computer Science 105 Computational Sprinting Although transistor density continues to increase, power density is increasing each chip generation. Particularly in mobile devices, which have limited cooling options, these trends lead to "dark silicon" in which sustained chip performance is limited primarily by power rather than area. However, many mobile applications do not demand sustained performance; rather they comprise short bursts of computation in response to sporadic user activity. To improve responsiveness for such applications, this talk describes "computational sprinting ", in which the system activates otherwise powered-down cores for brief bursts of intense parallel computation. During a sprint the chip temporarily exceeds its sustainable thermal power budget to provide instantaneous throughput, after which the chip must return to nominal operation to cool down. To enable longer sprints , we describe a thermal design that incorporates phase-change materials to provide thermal capacitance. Results indicate that sprinting can both provide large improvements in responsiveness while actually saving energy due to race-to-idle effects. Thomas Wenisch is an Associate Professor of Computer Science and Engineering at the University of Michigan, specializing in computer architecture. His prior research includes memory streaming for commercial server applications, multiprocessor memory systems, memory disaggregation, and rigorous sampling-based performance evaluation methodologies. His ongoing work focuses on computational sprinting , server and data center architectures, programming models for byte-addressable NVRAM, and architectures to enable hand-held 3D ultrasound . Wenisch received the NSF CAREER award in 2009. Prior to his academic career, Wenisch was a software developer at American Power Conversion, where he worked on data center thermal topology estimation. He received his Ph.D. in Electrical and Computer Engineering from Carnegie Mellon University.

1 0

Colloquium Speaker Thomas Wenisch, Monday, Dec 8, 4:30pm
by Nicole E. Wagenblast 07 Dec '14

07 Dec '14

Colloquium Speaker Thomas Wenisch, University of Michigan Monday, December 8, 4:30pm Computer Science 105 Computational Sprinting Although transistor density continues to increase, power density is increasing each chip generation. Particularly in mobile devices, which have limited cooling options, these trends lead to "dark silicon" in which sustained chip performance is limited primarily by power rather than area. However, many mobile applications do not demand sustained performance; rather they comprise short bursts of computation in response to sporadic user activity. To improve responsiveness for such applications, this talk describes "computational sprinting ", in which the system activates otherwise powered-down cores for brief bursts of intense parallel computation. During a sprint the chip temporarily exceeds its sustainable thermal power budget to provide instantaneous throughput, after which the chip must return to nominal operation to cool down. To enable longer sprints , we describe a thermal design that incorporates phase-change materials to provide thermal capacitance. Results indicate that sprinting can both provide large improvements in responsiveness while actually saving energy due to race-to-idle effects. Thomas Wenisch is an Associate Professor of Computer Science and Engineering at the University of Michigan, specializing in computer architecture. His prior research includes memory streaming for commercial server applications, multiprocessor memory systems, memory disaggregation, and rigorous sampling-based performance evaluation methodologies. His ongoing work focuses on computational sprinting , server and data center architectures, programming models for byte-addressable NVRAM, and architectures to enable hand-held 3D ultrasound . Wenisch received the NSF CAREER award in 2009. Prior to his academic career, Wenisch was a software developer at American Power Conversion, where he worked on data center thermal topology estimation. He received his Ph.D. in Electrical and Computer Engineering from Carnegie Mellon University.

1 0

Colloquium speaker Kristen Grauman, Today, 4:30pm
by Nicole E. Wagenblast 03 Dec '14

03 Dec '14

Colloquium Speaker Kristen Grauman, University of Texas at Austin Wednesday, December 3, 4:30pm Computer Science 105 Searching and Browsing Visual Data Widespread visual sensors and unprecedented connectivity have left us awash with visual data---from online photo collections, home videos, news footage, medical images, or surveillance feeds. How can we efficiently browse image and video collections based on semantically meaningful criteria? How can we bring order to the data, beyond manually defined keyword tags? I will present work exploring these questions in the context of interactive visual search and summarization. In particular, I’ll first introduce attribute representations that connect visual properties to human describable terms. I’ll show how these attributes enable both fine-grained content-based retrieval as well as new forms of human supervision for recognition problems. Then, I’ll overview our recent work on video summarization, where the goal is to automatically transform a long video into a short one. Using videos captured with egocentric wearable cameras, we’ll see how hours of data can be distilled to a succinct visual storyboard that is understandable in just moments. Together, these ideas are promising steps towards widening the channel of communication between humans and computer vision algorithms, which is critical to facilitate efficient browsing of large-scale image and video collections. This is work done with Adriana Kovashka, Yong Jae Lee, Devi Parikh, Lu Zheng, Bo Xiong, and Dinesh Jayaraman. Kristen Grauman is an Associate Professor in the Department of Computer Science at the University of Texas at Austin. Her research in computer vision and machine learning focuses on visual search and object recognition. Before joining UT-Austin in 2007, she received her Ph.D. in the EECS department at MIT, in the Computer Science and Artificial Intelligence Laboratory. She is an Alfred P. Sloan Research Fellow and Microsoft Research New Faculty Fellow, a recipient of NSF CAREER and ONR Young Investigator awards, the Regents' Outstanding Teaching Award from the University of Texas System in 2012, the PAMI Young Researcher Award in 2013, the 2013 Computers and Thought Award from the International Joint Conference on Artificial Intelligence, and a Presidential Early Career Award for Scientists and Engineers (PECASE) in 2013. She and her collaborators were recognized with the CVPR Best Student Paper Award in 2008 for their work on hashing algorithms for large-scale image retrieval, and the Marr Best Paper Prize at ICCV in 2011 for their work on modeling relative visual attributes.

1 0

Colloquium Speaker Thomas Wenisch, Monday, Dec 8, 4:30pm
by Nicole E. Wagenblast 02 Dec '14

02 Dec '14

Colloquium Speaker Thomas Wenisch, University of Michigan Monday, December 8, 4:30pm Computer Science 105 Computational Sprinting Although transistor density continues to increase, power density is increasing each chip generation. Particularly in mobile devices, which have limited cooling options, these trends lead to "dark silicon" in which sustained chip performance is limited primarily by power rather than area. However, many mobile applications do not demand sustained performance; rather they comprise short bursts of computation in response to sporadic user activity. To improve responsiveness for such applications, this talk describes "computational sprinting ", in which the system activates otherwise powered-down cores for brief bursts of intense parallel computation. During a sprint the chip temporarily exceeds its sustainable thermal power budget to provide instantaneous throughput, after which the chip must return to nominal operation to cool down. To enable longer sprints , we describe a thermal design that incorporates phase-change materials to provide thermal capacitance. Results indicate that sprinting can both provide large improvements in responsiveness while actually saving energy due to race-to-idle effects. Thomas Wenisch is an Associate Professor of Computer Science and Engineering at the University of Michigan, specializing in computer architecture. His prior research includes memory streaming for commercial server applications, multiprocessor memory systems, memory disaggregation, and rigorous sampling-based performance evaluation methodologies. His ongoing work focuses on computational sprinting , server and data center architectures, programming models for byte-addressable NVRAM, and architectures to enable hand-held 3D ultrasound . Wenisch received the NSF CAREER award in 2009. Prior to his academic career, Wenisch was a software developer at American Power Conversion, where he worked on data center thermal topology estimation. He received his Ph.D. in Electrical and Computer Engineering from Carnegie Mellon University.

1 0

Colloquium speaker Kristen Grauman, Wed Dec 3, 4:30pm
by Nicole E. Wagenblast 02 Dec '14

02 Dec '14

Colloquium Speaker Kristen Grauman, University of Texas at Austin Wednesday, December 3, 4:30pm Computer Science 105 Searching and Browsing Visual Data Widespread visual sensors and unprecedented connectivity have left us awash with visual data---from online photo collections, home videos, news footage, medical images, or surveillance feeds. How can we efficiently browse image and video collections based on semantically meaningful criteria? How can we bring order to the data, beyond manually defined keyword tags? I will present work exploring these questions in the context of interactive visual search and summarization. In particular, I’ll first introduce attribute representations that connect visual properties to human describable terms. I’ll show how these attributes enable both fine-grained content-based retrieval as well as new forms of human supervision for recognition problems. Then, I’ll overview our recent work on video summarization, where the goal is to automatically transform a long video into a short one. Using videos captured with egocentric wearable cameras, we’ll see how hours of data can be distilled to a succinct visual storyboard that is understandable in just moments. Together, these ideas are promising steps towards widening the channel of communication between humans and computer vision algorithms, which is critical to facilitate efficient browsing of large-scale image and video collections. This is work done with Adriana Kovashka, Yong Jae Lee, Devi Parikh, Lu Zheng, Bo Xiong, and Dinesh Jayaraman. Kristen Grauman is an Associate Professor in the Department of Computer Science at the University of Texas at Austin. Her research in computer vision and machine learning focuses on visual search and object recognition. Before joining UT-Austin in 2007, she received her Ph.D. in the EECS department at MIT, in the Computer Science and Artificial Intelligence Laboratory. She is an Alfred P. Sloan Research Fellow and Microsoft Research New Faculty Fellow, a recipient of NSF CAREER and ONR Young Investigator awards, the Regents' Outstanding Teaching Award from the University of Texas System in 2012, the PAMI Young Researcher Award in 2013, the 2013 Computers and Thought Award from the International Joint Conference on Artificial Intelligence, and a Presidential Early Career Award for Scientists and Engineers (PECASE) in 2013. She and her collaborators were recognized with the CVPR Best Student Paper Award in 2008 for their work on hashing algorithms for large-scale image retrieval, and the Marr Best Paper Prize at ICCV in 2011 for their work on modeling relative visual attributes.

1 0

Cole Schlesinger will present his Pre-FPO Tuesday, December 9 in CS 402 at 10am (Title: Abstractions for Software-defined Networks)
by Nicki Gotsis 02 Dec '14

02 Dec '14

Cole Schlesinger will present his Pre-FPO Tuesday, December 9, 2014 in CS 402 at 10am. The member of his committee are: Dave Walker (advisor), Aarti Gupta and Arjun Guha (readers), Nick Feamster and Jen Rexford (non-readers). Everyone is invited to attend his talk. The talk title and abstract follow below: Title: Abstractions for Software-defined Networks Abstract: In a Software-Defined Network (SDN), a central, computationally powerful controller manages a set of distributed, computationally simple switches. The controller computes a policy describing how each switch should route packets and populates packet-processing tables on each switch with rules to enact the routing policy. As network conditions change, the controller continues to add and remove rules from switches to adjust the policy as needed. My research focuses on defining new, high-level languages and abstractions for expressing network policy; defining technical mechanisms for reasoning formally about those abstractions; and defining algorithms for compiling high-level abstractions into available low-level mechanisms. In the first part of this talk, I present NetKAT, a language for expressing network-wide packet-forwarding policies. NetKAT comes equipped with a denotational semantics and a sound and complete equational theory drawn from a well-studied mathematical structure called a Kleene algebra with tests (KAT). The equational theory has many practical applications, such as syntactic techniques for checking reachability and proving non-interference properties that ensure isolation between policies. Here, I focus on applying the equational theory to guide and establish the correctness of a compilation algorithm targeting switches that expose simple, single-table packet-processing pipelines. In the second part of this talk, I show how to extend the switch-specific fragment of NetKAT with operators for writing and deploying network policies to switches with multiple-table, reconfigurable packet-processing pipelines. The language, called Concurrent NetCore, allows a network programmer to specify a virtual packet-processing pipeline, an abstraction that captures the physical resources required by the controller and also hides the details of the (potentially varied) pipelines of switches in the network. I also show how to use Concurrent NetCore to model the pipelines of three switching architectures and present compilation algorithms for mapping arbitrary virtual pipelines---and arbitrary packet-processing policies---to the physical pipelines. Along the way, I explore metatheoretical properties of the language and present correctness proofs for the compilation algorithms. Together, these abstractions and compilation algorithms pave the way for easier fine-grained network management. This talk draws on joint work with Carolyn Jane Anderson, Nate Foster, Michael Greenberg, Arjun Guha, Jean-Baptiste Jeannin, Dexter Kozen, and David Walker. Thanks! Cole

1 0

Colloquium speaker Edoardo Airoldi- Today, 4:30pm
by Nicole E. Wagenblast 02 Dec '14

02 Dec '14

Colloquium Speaker Edoardo Airoldi, Harvard University Tuesday, December 2, 4:30pm Computer Science 105 Statistical and machine learning challenges in the analysis of large networks Network data --- i.e., collections of measurements on pairs, or tuples, of units in a population of interest --- are ubiquitous nowadays in a wide range of machine learning applications, from molecular biology to marketing on social media platforms. Surprisingly, assumptions underlying popular statistical methods are often untenable in the presence of network data. Established machine learning algorithms often break when dealing with combinatorial structure. And the classical notions of variability, sample size and ignorability take unintended connotations. These failures open to door to a number of technical challenges, and to opportunities for introducing new fundamental ideas and for developing new insights. In this talk, I will discuss open statistical and machine learning problems that arise when dealing with large networks, mostly focusing on modeling and inferential issues, and provide an overview of key technical ideas and recent results and trends. Edoardo M. Airoldi is an Associate Professor of Statistics at Harvard University, where he leads the Harvard Laboratory for Applied Statistical Methodology. He holds a holds Ph.D. in Computer Science and an M.Sc. in Statistics from Carnegie Mellon University, and a B.Sc. in Mathematical Statistics and Economics from Bocconi University. His current research focuses on statistical theory and methods for designing and analyzing experiments in the presence of network interference, and on inferential issues that arise in models of network data. He works on applications in molecular biology and proteomics, and in social media analytics and marketing. Airoldi is the recipient several research awards including the ONR Young Investigator Award, the NSF CAREER Award, and the Alfred P. Sloan Research Fellowship, and has received several outstanding paper awards including the Thomas R. Ten Have Award for his work on causal inference, and the John Van Ryzin Award for his work in biology. He has recently advised the Obama for America 2012 campaign on their social media efforts, and serves as a technical advisor at Nanigans and Maxpoint.

1 0

Colloquium speaker Kristen Grauman, Wed Dec 3, 4:30pm
by Nicole E. Wagenblast 01 Dec '14

01 Dec '14

C olloquium Speaker Kristen Grauman, University of Texas at Austin Wednesday, December 3, 4:30pm Computer Science 105 Searching and Browsing Visual Data Widespread visual sensors and unprecedented connectivity have left us awash with visual data---from online photo collections, home videos, news footage, medical images, or surveillance feeds. How can we efficiently browse image and video collections based on semantically meaningful criteria? How can we bring order to the data, beyond manually defined keyword tags? I will present work exploring these questions in the context of interactive visual search and summarization. In particular, I’ll first introduce attribute representations that connect visual properties to human describable terms. I’ll show how these attributes enable both fine-grained content-based retrieval as well as new forms of human supervision for recognition problems. Then, I’ll overview our recent work on video summarization, where the goal is to automatically transform a long video into a short one. Using videos captured with egocentric wearable cameras, we’ll see how hours of data can be distilled to a succinct visual storyboard that is understandable in just moments. Together, these ideas are promising steps towards widening the channel of communication between humans and computer vision algorithms, which is critical to facilitate efficient browsing of large-scale image and video collections. This is work done with Adriana Kovashka, Yong Jae Lee, Devi Parikh, Lu Zheng, Bo Xiong, and Dinesh Jayaraman. Kristen Grauman is an Associate Professor in the Department of Computer Science at the University of Texas at Austin. Her research in computer vision and machine learning focuses on visual search and object recognition. Before joining UT-Austin in 2007, she received her Ph.D. in the EECS department at MIT, in the Computer Science and Artificial Intelligence Laboratory. She is an Alfred P. Sloan Research Fellow and Microsoft Research New Faculty Fellow, a recipient of NSF CAREER and ONR Young Investigator awards, the Regents' Outstanding Teaching Award from the University of Texas System in 2012, the PAMI Young Researcher Award in 2013, the 2013 Computers and Thought Award from the International Joint Conference on Artificial Intelligence, and a Presidential Early Career Award for Scientists and Engineers (PECASE) in 2013. She and her collaborators were recognized with the CVPR Best Student Paper Award in 2008 for their work on hashing algorithms for large-scale image retrieval, and the Marr Best Paper Prize at ICCV in 2011 for their work on modeling relative visual attributes.

1 0

Colloquium speaker Edoardo Airoldi- Tuesday, Dec 2, 4:30pm
by Nicole E. Wagenblast 01 Dec '14

01 Dec '14

Colloquium Speaker Edoardo Airoldi, Harvard University Tuesday, December 2, 4:30pm Computer Science 105 Statistical and machine learning challenges in the analysis of large networks Network data --- i.e., collections of measurements on pairs, or tuples, of units in a population of interest --- are ubiquitous nowadays in a wide range of machine learning applications, from molecular biology to marketing on social media platforms. Surprisingly, assumptions underlying popular statistical methods are often untenable in the presence of network data. Established machine learning algorithms often break when dealing with combinatorial structure. And the classical notions of variability, sample size and ignorability take unintended connotations. These failures open to door to a number of technical challenges, and to opportunities for introducing new fundamental ideas and for developing new insights. In this talk, I will discuss open statistical and machine learning problems that arise when dealing with large networks, mostly focusing on modeling and inferential issues, and provide an overview of key technical ideas and recent results and trends. Edoardo M. Airoldi is an Associate Professor of Statistics at Harvard University, where he leads the Harvard Laboratory for Applied Statistical Methodology. He holds a holds Ph.D. in Computer Science and an M.Sc. in Statistics from Carnegie Mellon University, and a B.Sc. in Mathematical Statistics and Economics from Bocconi University. His current research focuses on statistical theory and methods for designing and analyzing experiments in the presence of network interference, and on inferential issues that arise in models of network data. He works on applications in molecular biology and proteomics, and in social media analytics and marketing. Airoldi is the recipient several research awards including the ONR Young Investigator Award, the NSF CAREER Award, and the Alfred P. Sloan Research Fellowship, and has received several outstanding paper awards including the Thomas R. Ten Have Award for his work on causal inference, and the John Van Ryzin Award for his work in biology. He has recently advised the Obama for America 2012 campaign on their social media efforts, and serves as a technical advisor at Nanigans and Maxpoint.

1 0

Peng Sun's Pre FPO: Monday, December 8 at 10am in CS 402, Title: "Network Management Architecture for Delivering Cloud Services"
by Nicki Gotsis 01 Dec '14

01 Dec '14

Peng Sun will present his Pre FPO on Monday, December 8 at 10am in CS 402 His committee members are: Jennifer Rexford (advisor), Dave Walker and Nick Feamster (readers), Margaret Martonosi and Kai Li (non-readers) Everyone is invited to attend his talk. The talk title and abstract follow below: Title: Network Management Architecture for Delivering Cloud Services Abstract: Efficient and reliable networks are critical to good end-to-end performance of cloud services. In both the datacenters of cloud providers and the ISPs connecting clients, many network management applications were built to improve efficiency and reliability. However, the traditional way of building management applications is not sustainable, relying on heavy human involvement. Software-defined networking (SDN) promises to simplify programming management applications. Yet SDN's current scope is limited to routing control on network devices within a single domain. My research works explore expanding the concept of SDN for better end-to-end performance of cloud services. In the exploration, I take a research approach of achieving and balancing three objectives: 1) design simple and expressive programming abstraction, 2) build efficient and scalable system for execution, and 3) deploy and run in real world. Using this research approach, I have explored three directions. The first direction is deep into network devices to make infrastructure easy to program. The network operating system I built, called Statesman, introduces a three-view-of-network-state abstraction for programming a wide range of management applications. Besides the abstraction, Statesman enables automatic and safe operation of multiple applications in parallel. Published in SIGCOMM'14, Statesman has been deployed and operational worldwide in Microsoft Azure for over one year. The second direction is expanding into servers in datacenters to leverage the fine-grained knowledge of cloud services in the server OS. The system, called HONE, offers a three-stage programming framework to simplify traffic management. HONE's runtime executes the programs across servers and network devices in an efficient and scalable fashion. HONE has been integrated into products for Verizon Business Cloud, and appeared in JNSM July'14. The final direction is reaching out to ISPs for better delivery of cloud services. This work-in-progress aims at a direct, fine-grained, and self-contained solution to inbound traffic engineering (TE) for enterprises. The solution employs three tiers of abstractions to simplify the inbound-TE policies, and it scales up with a distributed enforcement system to control the inbound ISPs for traffic flows. This ongoing work is in collaboration with OIT on our campus network.

1 0