April 2018 - talks - lists.cs.princeton.edu

Reminder: Ben Lee, tomorrow at 4:30pm
by Emily C. Lawrence 30 Apr '18

30 Apr '18

Ben Lee from Duke University Tuesday, May 1, 2018 at 4:30 p.m. Maeder Hall auditorium Title: Datacenters and Energy Efficiency: A Game-Theoretic Perspective Abstract: Sharing datacenter hardware improves energy efficiency, but whether strategic users participate in consolidated systems depends on management policies. Users who dislike allocations may refuse to participate and deploy private, less-efficient systems. We rethink systems management, drawing on game theory to model strategic behavior and incentivize participation. We illustrate this perspective for two fundamental challenges in datacenters. For power delivery, we design sprinting games to produce equilibria in which users selfishly draw power for performance boosts yet avoid oversubscribing the shared supply. For resource allocation, we use Cobb-Douglas utility functions to produce fair allocations that incentivize users to share cache and memory. These solutions provide foundations for rigorously managing systems shared by strategic, competitive participants. Biography: Benjamin Lee is an Associate Professor of Electrical and Computer Engineering at Duke University. He received his B.S. from the University of California at Berkeley (2004), his Ph.D. from Harvard University (2008), and his post-doctorate from Stanford University (2010). He has held visiting positions at Microsoft Research, Intel Labs, and Lawrence Livermore National Lab. Dr. Lee’s research interests include computer architecture, energy efficiency, and security / privacy. He pursues these interests by building interdisciplinary links to statistical inference and algorithmic economics. His research has been recognized by IEEE Micro Top Picks (4x), Communications of the ACM Research Highlights (3x), as well as paper honors from the ASPLOS, HPCA, MICRO, and SC conferences. He received the NSF Computing Innovation Fellowship, NSF CAREER Award, and Google Faculty Research Award.

1 0

Colloquium Speaker-Bo Wang, TODAY at 12:30
by Emily C. Lawrence 27 Apr '18

27 Apr '18

Bo Wang, Stanford University Friday, April 27, 2018 - 12:30pm Computer Science - Room 105 Host: Mona Singh "Machine Learning Algorithms for Exploiting Spectral Structures of Biological Networks" Networks are ubiquitous in biology where they encode connectivity patterns at all scales of organization, from populations to a single cell. How to extract and understand non-trivial topological features and structures inherent in the networks is critical to understanding interactions within complicated biological systems. In this talk, I will introduce recent developments of machine learning algorithms that exploit spectral structures of networks for a wide range of biological applications, ranging from single-cell analysis to function prediction on protein-protein interaction networks. Specifically, I will first present a new method named SIMLR, combining both low-rank spectral regularization and multiple-kernel learning, to construct cell networks for sparse noisy single-cell RNA- seq data. The learned cell networks will enable effective dimension reduction, clustering and visualization. Second, I will discuss a novel method, Network Enhancement (NE), that aims to de-noise complex networks such as protein-protein interaction networks without corrupting spectral structures of the networks, therefore improving function prediction accuracy. Last, I will also briefly introduce recent advances where deep convolutional neural network is applied on biological networks (e.g., drug-target network) as a first-order spectral approximation of network structures. Bio: Bo Wang is a recent PhD graduate in Computer Science at Stanford University, an IEEE and CVF Fellow and an NSF Graduate Research Fellow. His research focuses on machine learning (particularly deep learning) on many applications in computer vision (e.g., image segmentation, object detection) and computational biology (e.g., single-cell analysis, integrative cancer subtyping). Prior to Stanford, he received his master degree at University of Toronto, majoring in numerical analysis.

1 0

CORRECTION - Xinyi Fan will present her FPO on Friday, May 4, 2018 at 1:00 pm in CS302
by Barbara A. Mooring 26 Apr '18

26 Apr '18

Good aftenoon, Xinyi Fan will present her FPO on Friday, May 4, 2018 at 1:00 pm in CS302 Title: View and Path Planning for Scaling 3D Acquisition to Many Objects Committee: Szymon Rusinkiewicz (advisor, COS) Adam Finkelstein (nonreader, COS) Peter Ramadge (nonreader, ELE) Tom Funkhouser (reader, COS) Yuxin Chen (reader, ELE) Abstract: Demand for high-volume 3D scanning of real objects is rapidly growing in a wide range of applications, including quality-control for manufacturing, online retailing, entertainment with virtual reality, as well as archaeological documentation and reconstruction. Fully realizing the potential of 3D acquisition requires scanning large numbers of objects with high quality and at reasonable cost. Although mature technologies exist for high-fidelity 3D model acquisition, deploying them at scale continues to require non-trivial manual labor. This dissertation focuses on studying practical 3D acquisition for large numbers of objects. The problem is challenging, because it is hard to automatically find a proper set of scanner views that can not only completely cover the surface of multiple objects with different shapes, but also capture high fidelity surface model of the objects. Furthermore, it is non-trivial to position a 3D scanner at each of the desired views accurately and efficiently. We propose a prototype system for multi-object 3D acquisition, which allows non-expert users to scan large numbers of physical objects within a reasonable amount of time, and with greater ease. Our system uses novel planning algorithms to control a structured-light scanner mounted on a calibrated motorized positioning system. We demonstrate the ability of our prototype to safely, robustly, and automatically acquire 3D models for large collections of small objects. We propose an objective function for automated view and path planning, taking into account both accuracy and efficiency of the scanning system. We analyze different approaches to optimize for the objective, and discuss their performance and practicality. In addition, we address the problem of surface inaccessibility to further refine our multi-object 3D acquisition system. We explore solutions for improvement from both the hardware and software ends. A copy of the thesis is on display in the F-wing of Equad rnoon, Barbara A. Mooring Interim Graduate Coordinator Computer Science Department Princeton University

1 0

Xinyl Fan will present his FPO on Friday, May 4, 2018 at 1:00 pm in CS302
by Barbara A. Mooring 26 Apr '18

26 Apr '18

Xinyl Fan will present his FPO on Friday, May 4, 2018 at 1:00 pm in CS302 Title: View and Path Planning for Scaling 3D Acquisition to Many Objects Committee: Szymon Rusinkiewicz (advisor, COS) Adam Finkelstein (nonreader, COS) Peter Ramadge (nonreader, ELE) Tom Funkhouser (reader, COS) Yuxin Chen (reader, ELE) Abstract: Demand for high-volume 3D scanning of real objects is rapidly growing in a wide range of applications, including quality-control for manufacturing, online retailing, entertainment with virtual reality, as well as archaeological documentation and reconstruction. Fully realizing the potential of 3D acquisition requires scanning large numbers of objects with high quality and at reasonable cost. Although mature technologies exist for high-fidelity 3D model acquisition, deploying them at scale continues to require non-trivial manual labor. This dissertation focuses on studying practical 3D acquisition for large numbers of objects. The problem is challenging, because it is hard to automatically find a proper set of scanner views that can not only completely cover the surface of multiple objects with different shapes, but also capture high fidelity surface model of the objects. Furthermore, it is non-trivial to position a 3D scanner at each of the desired views accurately and efficiently. We propose a prototype system for multi-object 3D acquisition, which allows non-expert users to scan large numbers of physical objects within a reasonable amount of time, and with greater ease. Our system uses novel planning algorithms to control a structured-light scanner mounted on a calibrated motorized positioning system. We demonstrate the ability of our prototype to safely, robustly, and automatically acquire 3D models for large collections of small objects. We propose an objective function for automated view and path planning, taking into account both accuracy and efficiency of the scanning system. We analyze different approaches to optimize for the objective, and discuss their performance and practicality. In addition, we address the problem of surface inaccessibility to further refine our multi-object 3D acquisition system. We explore solutions for improvement from both the hardware and software ends. A copy of the thesis is on display in the F-wing of Equad Barbara A. Mooring Interim Graduate Coordinator Computer Science Department Princeton University

1 0

Reminder: Colloquium Speaker-Bo Wang, tomorrow at 12:30
by Emily C. Lawrence 26 Apr '18

26 Apr '18

Bo Wang, Stanford University Friday, April 27, 2018 - 12:30pm Computer Science - Room 105 Host: Mona Singh "Machine Learning Algorithms for Exploiting Spectral Structures of Biological Networks" Networks are ubiquitous in biology where they encode connectivity patterns at all scales of organization, from populations to a single cell. How to extract and understand non-trivial topological features and structures inherent in the networks is critical to understanding interactions within complicated biological systems. In this talk, I will introduce recent developments of machine learning algorithms that exploit spectral structures of networks for a wide range of biological applications, ranging from single-cell analysis to function prediction on protein-protein interaction networks. Specifically, I will first present a new method named SIMLR, combining both low-rank spectral regularization and multiple-kernel learning, to construct cell networks for sparse noisy single-cell RNA- seq data. The learned cell networks will enable effective dimension reduction, clustering and visualization. Second, I will discuss a novel method, Network Enhancement (NE), that aims to de-noise complex networks such as protein-protein interaction networks without corrupting spectral structures of the networks, therefore improving function prediction accuracy. Last, I will also briefly introduce recent advances where deep convolutional neural network is applied on biological networks (e.g., drug-target network) as a first-order spectral approximation of network structures. Bio: Bo Wang is a recent PhD graduate in Computer Science at Stanford University, an IEEE and CVF Fellow and an NSF Graduate Research Fellow. His research focuses on machine learning (particularly deep learning) on many applications in computer vision (e.g., image segmentation, object detection) and computational biology (e.g., single-cell analysis, integrative cancer subtyping). Prior to Stanford, he received his master degree at University of Toronto, majoring in numerical analysis.

1 0

Joint PACM-CSML Colloquium: APR 30/David Blei (Rescheduled)
by Jennifer Rexford 24 Apr '18

24 Apr '18

> From: "Gina M. Holland" <gholland(a)Princeton.EDU> > Subject: NEXT: Joint PACM-CSML Colloquium: APR 30/David Blei (Rescheduled) > Date: April 24, 2018 at 10:35:12 AM EDT > To: "pacmdistribution(a)math.princeton.edu" <pacmdistribution(a)math.princeton.edu> > Cc: "Andrea E. La Bella" <alabella(a)Princeton.EDU> > > The next colloquium will be a Joint PACM-CSML Colloquium given by David Blei, Columbia University > > Title and abstract are below. > > ================= > DATE/LOCATION: Monday, April 30, 2018, 4:00 – 5:00 PM / 214 Fine Hall > > SPEAKER: David Blei, Columbia University > > TITLE: Scaling and Generalizing Bayesian Inference > > ABSTRACT: A core problem in statistics and machine learning is to approximate difficult-to-compute probability distributions. This problem is especially important in probabilistic modeling, which frames all inference about unknown quantities as a calculation about a conditional distribution. In this talk I review and discuss variational inference (VI), a method that approximates probability distributions through optimization. VI has been used in myriad applications in machine learning. It tends to be faster than more traditional methods, such as Markov chain Monte Carlo sampling. > > I discuss our recent research on VI. I first describe stochastic variational inference, an approximate inference algorithm for handling massive data sets, and demonstrate its application to probabilistic topic models of millions of articles. Then I discuss black box variational inference, a generic algorithm for approximating the posterior. Black box inference easily applies to many models but requires minimal mathematical work to implement. I will demonstrate black box inference on deep exponential families---a method for Bayesian deep learning---and describe how it enables powerful tools for probabilistic programming. > > ================= > >

1 0

Jeff Dean, TODAY at 4:30
by Emily C. Lawrence 24 Apr '18

24 Apr '18

Jeff Dean from Google Research Tuesday, April 24 , 2018 - 4:30pm Friend Center - Room 101 Host: Kai Li Event page: [ http://www.cs.princeton.edu/events/25679 | http://www.cs.princeton.edu/events/25679 ] "Deep Learning to Solve Challenging Problems" For the past six years, the Google Brain team has conducted research on difficult problems in artificial intelligence, on building large-scale computer systems for machine learning research, and, in collaboration with many teams at Google, on applying our research and systems to dozens of Google products. We have made significant progress in computer vision, speech recognition, language understanding, machine translation, healthcare, robotic control, and other areas. Our group has open-sourced the TensorFlow system, a widely popular system designed to easily express machine learning ideas, and to quickly train, evaluate and deploy machine learning systems. In this talk, I'll highlight some of the research and computer systems work we've done with an eye towards how it can be used to solve challenging problems. This talk describes joint work with many people at Google. Bio: Jeff Dean joined Google in 1999 and is currently a Google Senior Fellow in Google's Research Group, where he co-founded and leads the Google Brain team, Google's deep learning and artificial intelligence research team. He and his collaborators are working on systems for speech recognition, computer vision, language understanding, and various other machine learning tasks. He has co-designed/implemented many generations of Google's crawling, indexing, and query serving systems, and co-designed/implemented major pieces of Google's initial advertising and AdSense for Content systems. He is also a co-designer and co- implementor of Google's distributed computing infrastructure, including the MapReduce , BigTable and Spanner systems, protocol buffers, the open-source TensorFlow system for machine learning, and a variety of internal and external libraries and developer tools. Jeff received a Ph.D. in Computer Science from the University of Washington in 1996, working with Craig Chambers on whole-program optimization techniques for object-oriented languages. He received a B.S. in computer science & economics from the University of Minnesota in 1990. He is a member of the National Academy of Engineering, and of the American Academy of Arts and Sciences, a Fellow of the Association for Computing Machinery (ACM), a Fellow of the American Association for the Advancement of Sciences (AAAS), and a winner of the ACM Prize in Computing and the Mark Weiser Award. **A reception in the Friend Center Upper Atrium will follow immediately after the talk. ***This event will be LIVESTREAMED! Please visit this link to watch the talk. [ https://mediacentrallive.princeton.edu/ | https://mediacentrallive.princeton.edu/ ]

1 0

Colloquium Speaker-Umesh Vazirani, TODAY at 12:30
by Emily C. Lawrence 23 Apr '18

23 Apr '18

Umesh Vazirani from UC Berkeley Monday, April 23, 2018 - 12:30pm Computer Science - Room 105 Host: Sanjeev Arora “Quantum Supremacy” and the Complexity of Random Circuit Sampling A critical goal for the field of quantum computation is quantum supremacy -- a demonstration of any quantum computation that is prohibitively hard for classical computers. Besides dispelling any skepticism about the viability of quantum computers, quantum supremacy also provides a test of quantum theory in the realm of high complexity. A leading near-term candidate, put forth by the Google/UCSB team, is sampling from the probability distributions of randomly chosen quantum circuits, called Random Circuit Sampling (RCS). While RCS was defined with experimental realization in mind (the first results are expected later this year), we give the first complexity-theoretic evidence of classical hardness of RCS, placing it on par with the best theoretical proposals for supremacy. Specifically, we show that RCS satisfies an average-case hardness condition -- computing output probabilities of typical quantum circuits is as hard as computing them in the worst-case, and therefore #P-hard. Our reduction exploits the polynomial structure in the output amplitudes of random quantum circuits, enabled by the Feynman path integral. We also describe a new verification measure which in some formal sense maximizes the information gained from experimental samples. Based on joint work with Adam Bouland , Bill Fefferman and Chinmay Nirkhe .

1 0

Reminder: Distinguished Lecturer Jeff Dean, tomorrow at 4:30
by Emily C. Lawrence 23 Apr '18

23 Apr '18

Jeff Dean from Google Research Tuesday, April 24, 2018 - 4:30pm Friend Center - Room 101 Host: Kai Li Event page: [ http://www.cs.princeton.edu/events/25679 | http://www.cs.princeton.edu/events/25679 ] "Deep Learning to Solve Challenging Problems" For the past six years, the Google Brain team has conducted research on difficult problems in artificial intelligence, on building large-scale computer systems for machine learning research, and, in collaboration with many teams at Google, on applying our research and systems to dozens of Google products. We have made significant progress in computer vision, speech recognition, language understanding, machine translation, healthcare, robotic control, and other areas. Our group has open-sourced the TensorFlow system, a widely popular system designed to easily express machine learning ideas, and to quickly train, evaluate and deploy machine learning systems. In this talk, I'll highlight some of the research and computer systems work we've done with an eye towards how it can be used to solve challenging problems. This talk describes joint work with many people at Google. Bio: Jeff Dean joined Google in 1999 and is currently a Google Senior Fellow in Google's Research Group, where he co-founded and leads the Google Brain team, Google's deep learning and artificial intelligence research team. He and his collaborators are working on systems for speech recognition, computer vision, language understanding, and various other machine learning tasks. He has co-designed/implemented many generations of Google's crawling, indexing, and query serving systems, and co-designed/implemented major pieces of Google's initial advertising and AdSense for Content systems. He is also a co-designer and co-implementor of Google's distributed computing infrastructure, including the MapReduce, BigTable and Spanner systems, protocol buffers, the open-source TensorFlow system for machine learning, and a variety of internal and external libraries and developer tools. Jeff received a Ph.D. in Computer Science from the University of Washington in 1996, working with Craig Chambers on whole-program optimization techniques for object-oriented languages. He received a B.S. in computer science & economics from the University of Minnesota in 1990. He is a member of the National Academy of Engineering, and of the American Academy of Arts and Sciences, a Fellow of the Association for Computing Machinery (ACM), a Fellow of the American Association for the Advancement of Sciences (AAAS), and a winner of the ACM Prize in Computing and the Mark Weiser Award. **A reception in the Friend Center Upper Atrium will follow immediately after the talk.

1 0

Reminder: Umesh Vazirani, tomorrow at 12:30
by Emily C. Lawrence 22 Apr '18

22 Apr '18

Umesh Vazirani from UC Berkeley Monday, April 23, 2018 - 12:30pm Computer Science - Room 105 Host: Sanjeev Arora “Quantum Supremacy” and the Complexity of Random Circuit Sampling A critical goal for the field of quantum computation is quantum supremacy -- a demonstration of any quantum computation that is prohibitively hard for classical computers. Besides dispelling any skepticism about the viability of quantum computers, quantum supremacy also provides a test of quantum theory in the realm of high complexity. A leading near-term candidate, put forth by the Google/UCSB team, is sampling from the probability distributions of randomly chosen quantum circuits, called Random Circuit Sampling (RCS). While RCS was defined with experimental realization in mind (the first results are expected later this year), we give the first complexity-theoretic evidence of classical hardness of RCS, placing it on par with the best theoretical proposals for supremacy. Specifically, we show that RCS satisfies an average-case hardness condition -- computing output probabilities of typical quantum circuits is as hard as computing them in the worst-case, and therefore #P-hard. Our reduction exploits the polynomial structure in the output amplitudes of random quantum circuits, enabled by the Feynman path integral. We also describe a new verification measure which in some formal sense maximizes the information gained from experimental samples. Based on joint work with Adam Bouland, Bill Fefferman and Chinmay Nirkhe.

1 0