November 2014 - talks - lists.cs.princeton.edu

Colloquium Speaker Sham Kakade Monday, Dec 1, 4:30pm
by Nicole E. Wagenblast 30 Nov '14

30 Nov '14

Colloquium Speaker Sham Kakade, Microsoft Research, New England Monday, December 1, 4:30pm Computer Science 105 Some Algorithmic Challenges in Statistics Machine learning is seeing tremendous progress in its impact on society. Along with this progress comes an increasing role for both scalable algorithms and theoretical foundations; the hope being that the such progress can facilitate further breakthroughs on core AI problems. This will talk will survey recent progress and future challenges at the intersection of computer science and statistics, with a focus on three areas: How can we learn the interactions between observed variables, where there exist certain latent (or hidden) causes which help to explain the correlations in the observed data. Such settings where latent variable models have seen successes include document (or topic) modeling, hidden Markov models (say for modeling time series of acoustic signals), and discovering communities of individuals in social networks. The second is that of stochastic optimization. Many problems that arise in science and engineering are those in which we only have a stochastic approximation to the underlying problem at hand (e.g. linear regression or other problems where our objective function is a sample average). Such problems highlight some of the challenges we face at the interface of computer science and statistics: should we use a highly (numerically) accurate algorithm (with costly time and space requirements) or a crude stochastic approximation scheme like stochastic gradient descent (which is light on memory and simple to implement, yet has a poor convergence rate)? Finally, I will provide a brief discussion with regards to future challenges inspired by the impressive successes of deep learning. A recurring theme is that algorithmic advances can provide new practical techniques for statistical estimation. Sham Kakade is a principal research scientist scientist at Microsoft Research, New England. His research focus is on designing scalable and efficient algorithms for machine learning and artificial intelligence; he has worked (and has continued interests) in areas such as statistics, optimization, probability theory, algorithms, economics, and neuroscience. Previously, Dr. Kakade was an associate professor at the Department of Statistics, Wharton, University of Pennsylvania (from 2010-2012) and was an assistant professor at the Toyota Technological Institute at Chicago. Before this, he did a postdoc in the Computer and Information Science department at the University of Pennsylvania under the supervision of Michael Kearns. Dr. Kakade completed his PhD at the Gatsby Unit where his advisor was Peter Dayan. Before Gatsby, Dr. Kakade was an undergraduate at Caltech where he did his BS in physics.

1 0

Colloquium Speaker Sham Kakade Monday, Dec 1, 4:30pm
by Nicole E. Wagenblast 25 Nov '14

25 Nov '14

Colloquium Speaker Sham Kakade, Microsoft Research, New England Monday, December 1, 4:30pm Computer Science 105 Some Algorithmic Challenges in Statistics Machine learning is seeing tremendous progress in its impact on society. Along with this progress comes an increasing role for both scalable algorithms and theoretical foundations; the hope being that the such progress can facilitate further breakthroughs on core AI problems. This will talk will survey recent progress and future challenges at the intersection of computer science and statistics, with a focus on three areas: How can we learn the interactions between observed variables, where there exist certain latent (or hidden) causes which help to explain the correlations in the observed data. Such settings where latent variable models have seen successes include document (or topic) modeling, hidden Markov models (say for modeling time series of acoustic signals), and discovering communities of individuals in social networks. The second is that of stochastic optimization. Many problems that arise in science and engineering are those in which we only have a stochastic approximation to the underlying problem at hand (e.g. linear regression or other problems where our objective function is a sample average). Such problems highlight some of the challenges we face at the interface of computer science and statistics: should we use a highly (numerically) accurate algorithm (with costly time and space requirements) or a crude stochastic approximation scheme like stochastic gradient descent (which is light on memory and simple to implement, yet has a poor convergence rate)? Finally, I will provide a brief discussion with regards to future challenges inspired by the impressive successes of deep learning. A recurring theme is that algorithmic advances can provide new practical techniques for statistical estimation. Sham Kakade is a principal research scientist scientist at Microsoft Research, New England. His research focus is on designing scalable and efficient algorithms for machine learning and artificial intelligence; he has worked (and has continued interests) in areas such as statistics, optimization, probability theory, algorithms, economics, and neuroscience. Previously, Dr. Kakade was an associate professor at the Department of Statistics, Wharton, University of Pennsylvania (from 2010-2012) and was an assistant professor at the Toyota Technological Institute at Chicago. Before this, he did a postdoc in the Computer and Information Science department at the University of Pennsylvania under the supervision of Michael Kearns. Dr. Kakade completed his PhD at the Gatsby Unit where his advisor was Peter Dayan. Before Gatsby, Dr. Kakade was an undergraduate at Caltech where he did his BS in physics.

1 0

Thiago Pereira will present his pre-FPO 11/25/14 at 10am in rm 402
by Nicki Gotsis 24 Nov '14

24 Nov '14

Thiago Pereira will present his Pre-FPO on Tuesday, November 25, 2014 at 10am in rm 402. The members of his committee is Szymon Rusinkiewicz (advisor), Wojciech Matusik and Tom Funkhouser (readers), Adam Finkelstein and Jianxiong Xiao (non readers). Everyone is invited to attend his talk. The talk title and abstract follow below: "New methods for the computational fabrication of appearance" Computational fabrication of appearance is an exciting and challenging new research direction. Its objective is creating objects that interact with light in a manner prescribed in a computational model. It can have great impact in a number of different fields including product prototyping and design, medical applications such as realistic prosthesis, lighting interior environments in architecture and watermarks in security. In this thesis, we propose three works on appearance fabrication: an appearance similarity metric, a light routing algorithm and new reflectance fabrication process. First, recent spatially varying reflectance (svBRDF) printing systems can reproduce an input document as a combination of matte, glossy and metallic inks. Due to the limited number of inks, this reproduction process incurs some distortion. In this work, we present an svBRDF gamut mapping algorithm that minimizes distortions in the angular and spatial domains. To preserve a material's perceived variation with lighting and view, we introduce an improved BRDF similarity metric that builds on both experimental results on reflectance perception and on the statistics of natural lighting environments. Our experiments show better preservation of object color and highlights, as validated quantitatively as well as through a perceptual study. As for the spatial domain, we show how to adapt traditional color gamut mapping methods to svBRDFs. Our solution takes into account the contrast between regions, achieving better preservation of textures and edges. Second, despite recent interest in digital fabrication, there are still few algorithms that provide control over how light propagates inside a solid object. Existing methods either work only on the surface or restrict themselves to light diffusion in volumes. We use multi-material 3D printing to fabricate objects with embedded optical fibers, exploiting total internal reflection to guide light inside an object. We introduce automatic fiber design algorithms together with new manufacturing techniques to route light between two arbitrary surfaces. Our implicit algorithm optimizes light transmission by minimizing fiber curvature and maximizing fiber separation while respecting constraints such as fiber arrival angle. We also discuss the influence of different printable materials and fiber geometry on light propagation in the volume and the light angular distribution when exiting the fiber. Our methods enable new applications such as surface displays of arbitrary shape, touch-based painting of surfaces and sensing a hemispherical light distribution in a single shot. Third, existing BRDF fabrication methods still have many limitations such as: symmetry of light scattering or restriction to small scattering angles. We propose to use magnetic reflective pigments such that we can control their orientation by applying a magnetic field. We use dynamic magnetic fields for controlling not only off-specular lobes direction but also lobe width and anisotropy. We show how this magnetic control can be coupled with a projector to achieve new spatially-varying anisotropic BRDFs.

1 0

Fwd: EE Seminar - December 1, 2014, 4:30 pm, E-Quad B205 - Prof. Johan Pouwelse
by Jennifer Rexford 24 Nov '14

24 Nov '14

> From: "Lori A. Bailey" <lbailey(a)Princeton.EDU> > Date: November 24, 2014 at 12:42:33 PM EST > To: ee-seminar(a)Princeton.EDU > Subject: EE Seminar - December 1, 2014, 4:30 pm, E-Quad B205 - Prof. Johan Pouwelse > Reply-To: "Lori A. Bailey" <lbailey(a)Princeton.EDU> > > > Speaker: Prof. Johan Pouwelse > Delft University of Technology > Title: Anonymous HD Video Streaming and Reputations > Date: Monday, December 1, 2014 > Time: 4:30 p.m. > Room: E-Quad, Room B205 > Host: Prof. Prateek Mittal > > Abstract: Privacy is non-existent for 99% of all web traffic. We provide Tor-like privacy protection for the most demanding class of Internet traffic: HD-quality video streams. To date, no viable solution has ever been presented for privacy-enhanced streaming. The challenge is to scale to one billion users, the current usage level of Youtube. > > Youtube tracks what you like, when you view it, where you watch it and what your friends are viewing. Scaling privacy-enhancing technology while retaining security has proving to be insurmountable. Many proposals exists, but no system has any significant usage (e.g. 1+ million users) or has even been fully implemented. We present the first functional fully self-organising darknet for HD-quality video streaming. The key building block is our BarterCast reputation system, indicating which users help others by relaying their encrypted traffic. Our darknet, called Tribler, evolved over a period of nine years, has a thriving user community, comes with proper unit tests, coding tutorials, and is designed to be fault-tolerant by not relying on any central server, tracker or website. > > Biography: Dr. ir. J.A. Pouwelse is an associate professor at Delft University of Technology, specialized in self-organizing systems and privacy-enhancing technology. His research group conducts experimental research in the field of self-organizing systems, reputation systems, crowdsourcing, onion routing and peer-to-peer technology. The methodology of the group is proving the validity of novel scientific ideas with thousands of real-world Internet users. Our Tribler software is installed by 1.6 million users, it serves as a living laboratory and testing ground for next-generation self-organizing systems. Previously Dr. Pouwelse delivered a statement for the FTC in Washington, was a visiting scientist at MIT, and spent several summers at Harvard to study mechanisms for cooperation. > > > > >

1 0

Fwd: EE Seminar -TODAY at 4:30 pm, E-Quad B205 - Prof. Lakshminarayanan Subramanian4:30 pm, E-Quad B205 - Prof. Lakshminarayanan Subramanian
by Jennifer Rexford 24 Nov '14

24 Nov '14

> > SEMINAR TODAY > > > Speaker: Prof. Lakshminarayanan Subramanian > > New York University > > Title: Societal Network Science: Where Societal Networks meets Data Science > > Date: Monday, November 24, 2014 > > Time: 4:30 p.m. > > Room: E-Quad, Room B205 > > Host: Prof. Prateek Mittal > > > > Abstract: This talk focuses on a new emerging area of research, which we term "Societal Network Science", where the goal is to understand important societal, social and human indicators based on the network characteristics of how humans, physical and digital entities interact. Societal networks aim to capture meaningful human interaction patterns and the potential causal relationships between these interactions. In this talk, we will outline our recent results on three related but seemingly disparate problems in societal network science: disease surveillance, road traffic networks and event networks. In the space of disease surveillance, our group has worked on a 3-year long project to build a fine-grained dengue surveillance system in the country of Pakistan, where we can predict the incidence level of dengue using citizen driven data at a sub-city granularity, 14 days ahead of time. In the space of road traffic networks, we will illustrate how one can predict sudden traffic jams in NY city with high confidence 3 minutes before they happen; surprisingly, 3 minutes is sufficient to potentially avoid the jam too! In the space of event networks, we will show how one can use NLP techniques to learn networks of real world events from news corpora and how such event networks can be used in the prediction of real-world social phenomena like drought, price variations and disease outbreaks. Our longer term vision is to build an automated analytics engine for societal network science that can learn and infer socio-economic phenomena from highly diverse and noisy data sources. > > > > Biography: Lakshminarayanan Subramanian is an Associate Professor in the Courant Institute of Mathematical Sciences at NYU. His research interests are in the areas of networked systems, big data analytics and computing for development. He is a member of the NYU Systems group and co-established the Center for Technology and Economic Development (CTED) at NYU. He is also affiliated with the Center for Data Science and the Global Institute of Public Health (GIPH) at NYU. He is the co-founder of Entrupy Inc, a startup on surface fingerprinting. He completed his PhD from UC Berkeley and undergraduate studies from IIT-Madras. >

1 0

Colloquium Speaker Erik Sudderth, Today, 4:30pm
by Nicole E. Wagenblast 24 Nov '14

24 Nov '14

Colloquium Speaker Erik Sudderth, Brown University Monday, November 24, 2014 - 4:30pm Computer Science 105 Flexible, Reliable, and Scalable Nonparametric Learning Applications of statistical machine learning increasingly involve datasets with rich hierarchical, temporal, spatial, or relational structure. Bayesian nonparametric models offer the promise of effective learning from big datasets, but standard inference algorithms often fail in subtle and hard-to-diagnose ways. We explore this issue via variants of a popular and general model family, the hierarchical Dirichlet process. We propose a framework for "memoized" online optimization of variational learning objectives, which achieves computational scalability by processing local batches of data, while simultaneously adapting the global model structure in a coherent fashion. Using this approach, we build improved models of text, audio, image, and social network data. Erik B. Sudderth is an Assistant Professor in the Brown University Department of Computer Science. He received the Bachelor's degree (summa cum laude, 1999) in Electrical Engineering from the University of California, San Diego, and the Master's and Ph.D. degrees (2006) in EECS from the Massachusetts Institute of Technology. His research interests include probabilistic graphical models; nonparametric Bayesian methods; and applications of statistical machine learning in computer vision and the sciences. He received an NSF CAREER award in 2014, and in 2008 was named one of "AI's 10 to Watch" by IEEE Intelligent Systems Magazine.

1 0

Colloquium Speaker Erik Sudderth, Wed Nov 24, 4:30pm
by Nicole E. Wagenblast 23 Nov '14

23 Nov '14

Colloquium Speaker Erik Sudderth, Brown University Wednesday, November 24, 2014 - 4:30pm Computer Science 105 Flexible, Reliable, and Scalable Nonparametric Learning Applications of statistical machine learning increasingly involve datasets with rich hierarchical, temporal, spatial, or relational structure. Bayesian nonparametric models offer the promise of effective learning from big datasets, but standard inference algorithms often fail in subtle and hard-to-diagnose ways. We explore this issue via variants of a popular and general model family, the hierarchical Dirichlet process. We propose a framework for "memoized" online optimization of variational learning objectives, which achieves computational scalability by processing local batches of data, while simultaneously adapting the global model structure in a coherent fashion. Using this approach, we build improved models of text, audio, image, and social network data. Erik B. Sudderth is an Assistant Professor in the Brown University Department of Computer Science. He received the Bachelor's degree (summa cum laude, 1999) in Electrical Engineering from the University of California, San Diego, and the Master's and Ph.D. degrees (2006) in EECS from the Massachusetts Institute of Technology. His research interests include probabilistic graphical models; nonparametric Bayesian methods; and applications of statistical machine learning in computer vision and the sciences. He received an NSF CAREER award in 2014, and in 2008 was named one of "AI's 10 to Watch" by IEEE Intelligent Systems Magazine.

1 2

Colloquium Speaker Sebastian Burckhardt Today, 4:30pm
by Nicole E. Wagenblast 20 Nov '14

20 Nov '14

Colloquium Speaker Sebastian Burckhardt , from Microsoft Research Thursday, November 20, 2014 - 4:30pm Computer Science Small Auditorium (Room 105) Robust Abstractions for Replicated Shared State In the age of cloud-connected mobile devices, users want responsive apps that read and write shared data everywhere, at all times, even if network connections are slow or unavailable. Replication and eventual consistency, while able to deliver this experience, require us to face the complexity of asynchronous update propagation and conflict resolution. Our research goal is to find abstractions that encapsulate this complexity, in order to simplify the programming of distributed applications that are responsive, reactive, and collaborative. In this talk, we first discuss the general principles of eventual consistency. Then, we introduce our programming model, consisting of cloud types (for declarative type-based conflict resolution) and the GLUT model (an operational consistency model based on a global log of update transactions). Finally, we report on our practical experiences with supporting cloud types and GLUT in the TouchDevelop programming language and mobile development environment. Sebastian Burckhardt was born and raised in Basel, Switzerland, where he studied Mathematics at the local University. During an exchange year at Brandeis University, he discovered his affinity to Computer Science and immigrated to the United States. After a few years of industry experience at IBM, he returned to academia and earned his PhD in Computer Science at the University of Pennsylvania. Since then, he has worked as a researcher at Microsoft Research in Redmond. His general research interest is the study of programming models for of concurrent, parallel, and distributed systems. More specific interests include consistency models, concurrency testing, self-adjusting computation, and the concurrent revisions programming model.

1 0

Colloquium Speaker Sebastian Burckhardt Thurs Nov 20, 4:30pm
by Nicole E. Wagenblast 19 Nov '14

19 Nov '14

Colloquium Speaker Sebastian Burckhardt , from Microsoft Research Thursday, November 20, 2014 - 4:30pm Computer Science Small Auditorium (Room 105) Robust Abstractions for Replicated Shared State In the age of cloud-connected mobile devices, users want responsive apps that read and write shared data everywhere, at all times, even if network connections are slow or unavailable. Replication and eventual consistency, while able to deliver this experience, require us to face the complexity of asynchronous update propagation and conflict resolution. Our research goal is to find abstractions that encapsulate this complexity, in order to simplify the programming of distributed applications that are responsive, reactive, and collaborative. In this talk, we first discuss the general principles of eventual consistency. Then, we introduce our programming model, consisting of cloud types (for declarative type-based conflict resolution) and the GLUT model (an operational consistency model based on a global log of update transactions). Finally, we report on our practical experiences with supporting cloud types and GLUT in the TouchDevelop programming language and mobile development environment. Sebastian Burckhardt was born and raised in Basel, Switzerland, where he studied Mathematics at the local University. During an exchange year at Brandeis University, he discovered his affinity to Computer Science and immigrated to the United States. After a few years of industry experience at IBM, he returned to academia and earned his PhD in Computer Science at the University of Pennsylvania. Since then, he has worked as a researcher at Microsoft Research in Redmond. His general research interest is the study of programming models for of concurrent, parallel, and distributed systems. More specific interests include consistency models, concurrency testing, self-adjusting computation, and the concurrent revisions programming model.

1 0

Colloquium Speaker Emma Brunskill Today, 4:30pm
by Nicole E. Wagenblast 19 Nov '14

19 Nov '14

Colloquium Speaker Emma Brunskill , Carnegie Mellon University Wednesday, November 19, 2014 - 4:30pm Computer Science 105 Interactive ML for People: The Small Data Problem Consider an intelligent tutoring system or an autonomous decision support tool for a doctor. Though such systems may in aggregate have a huge amount of data, the data collected for a single individual is typically very small, and the policy space (of what to next teach a student or how to help treat a patient) is enormous. I will describe two machine learning efforts to tackle these small data challenges: learning across multiple tasks, and better use of previously collected task data, where tasks in both cases involve sequential stochastic decision processes (reinforcement learning and bandits). I will also present results of how one of these techniques allowed us to substantially increase engagement in an educational game to teach fractions. Emma Brunskill is an assistant professor in the computer science department at Carnegie Mellon University. She is also affiliated with the machine learning department at CMU. She works on reinforcement learning, focusing on applications that involve artificial agents interacting with people, such as intelligent tutoring systems. She is a Rhodes Scholar, Microsoft Faculty Fellow and NSF CAREER award recipient, and her work has received best paper nominations in Education Data Mining (2012, 2013) and CHI (2014).

1 0