Tuesday 13 September 2016

View the recording on Lecturecast (UCL login required) here

Can data science help reduce police violence and misconduct? Can it help prevent children from getting lead poisoning? Can it help cities better target limited resources to improve lives of citizens? We're all aware of the data science hype right now but turning this hype into any social impact takes effort. In this talk, I'll discuss lessons learned from our work at University of Chicago while working on dozens of data science projects over the past few years with non-profits and governments on high-impact public policy and social challenges. These lessons span from challenges these organizations face when trying to apply data science, to understanding how to effectively train and build cross-disciplinary teams to do practical data science, as well as what data science and social science research challenges need to be tackled, and what tools and techniques need to be developed in order to have a social and policy impact with data science.

Rayid is a reformed computer scientist and wanna-be social scientist, but mostly just wants to increase the use of data-driven approaches in solving large public policy and social challenges. Rayid is also passionate about teaching practical data science and started the Eric & Wendy Schmidt Data Science for Social Good Fellowship at UChicago that trains computer scientists, statisticians, and social scientists from around the world to work on data science problems with social impact. Before joining the University of Chicago, Rayid was the Chief Scientist of the Obama 2012 Election Campaign where he focused on data, analytics, and technology to target and influence voters, donors, and volunteers. Previously, Rayid was a Research Scientist and led the Machine Learning group at Accenture Labs. Rayid did his graduate work in Machine Learning at Carnegie Mellon University and is actively involved in organizing Data Science related conferences and workshops. In his ample free time, Rayid works with governments and non-profits to help them with their data, analytics and digital efforts and strategy.

Tuesday 26 July 2016

View the recording on Lecturecast (UCL login required) here

A common theme in program verification is establishing relationships between two runs of the same program or of different programs. Such relationships can be proved by semantical means, or with syntactic methods such as relational program logics and product constructions. Gilles shall present an overview of these methods and their applications to provable security, differential privacy, and secure implementations.

Gilles Barthe is a research professor at the IMDEA Software Institute. His research interests include logic, formal verification, programming languages, and security. His current work focuses on verification and synthesis methods for cryptography and differential privacy. He is a member of the editorial boards of the Journal of Automated Reasoning and Journal of Computer Security. He received a Ph.D. in Mathematics from the University of Manchester, UK, in 1993, and an Habilitation à diriger les recherches in Computer Science from the University of Nice, France, in 2004.

Thursday 5 November 2015

View the recording on Lecturecast (UCL login required) here

This is a story of transporting ideas from theoretical research in reasoning about programs into the fast-moving engineering culture of Facebook. The context is that I landed at Facebook in September of 2013, when we brought the Infer static analyser with us from the verification startup Monoidics. Infer is based on recent research in program analysis, which applied a relatively recent development in logics of programs, separation logic. Infer is deployed internally, running continuously to verify select properties of every code modification in Facebook's mobile apps; these include the main Facebook apps for Android and iOS, Facebook Messenger, Instagram, and other apps which are used by over a billion people in total. This talk describes our experience deploying verification technology inside Facebook, some the challenges we faced, lessons learned, and speculates on prospects for broader impact of verification technology.

Peter O'Hearn works as an Engineering Manager at Facebook with the Static Analysis Tools team, and as a Professor of Computer Science at UCL. His research has been in the broad areas of programming languages and logic, ranging from new logics and mathematical models to industrial applications of program proof. With John Reynolds he developed separation logic, a theory which opened up new practical possibilities for program proof. In 2009 he cofounded a software verification startup company, Monoidics Ltd, which was acquired by Facebook in 2013. The Facebook Infer program analyzer, recently open-sourced, runs on every modification to the code of Facebook's mobile apps, in a typical month issuing millions of calls to a custom separation logic theorem prover and catching hundreds of bugs before they reach production.

Wednesday 10 June 2015

View the recording on Lecturecast (UCL login required) here

The last decade has witnessed rapid advancements in computer vision systems, not just in the world of gaming, but in many aspects of everyday life from medical systems to augmented reality. Computer systems “that see” enable new forms of input, can track and identify people, can capture and model the physical world around us, and can be combined with other system capabilities such as conversational agents.  But the challenge in developing these systems is much more than technical. In this talk I explore the process of designing computer vision applications from a human perspective, and through our own attempts to build them for a variety of real world settings.  In doing so, I propose that such systems need to make their users aware of the differences between how computer systems and how people sense, perceive, analyse and respond to the world.  This has implications beyond computer vision to more general notions of “smart” systems in an era where artificial intelligence has again taken hold of our collective imagination.

Abigail Sellen is a Principal Researcher at Microsoft Research Cambridge where she manages the Human Experience & Design Group. Prior to Microsoft, she worked at Hewlett-Packard Labs, Rank Xerox EuroPARC, Apple Computer and Bell Northern Research. Abigail first became interested in Human-Computer Interaction through a summer internship at Apple while working on her doctorate in Cognitive Science with Don Norman.  She has since published extensively on many diverse topics including the book "The Myth of the Paperless Office" (with co-author Richard Harper). Alongside her honorary professorship at UCL, she is also a Fellow of the Royal Academy of Engineering, Fellow of the British Computer Society, and a member of the ACM SIGCHI Academy.

Thursday 22 January 2015

View the recording on Lecturecast (UCL login required) here

This talk will cover some of our recent work in extended topic models to serve as tools in text mining and NLP (and hopefully, later, in IR) when some semantic analysis is required.  In some sense our goals are akin to the use of Latent Semantic Analysis.  The basic theoretical/algorithmic tool we have for this is non-parametric Bayesian methods for reasoning on hierarchies of probability vectors. The concepts will be introduced but not the statistical detail. Then I'll present some of our KDD 2014 paper (Experiments with Non-parametric Topic Models), and some extended work such as "Bibliographic Analysis with the Citation Network Topic Model" (ACML 2014) and "Topic Segmentation with a Structured Topic Model" (NAACL 2013).  Various evaluations and comparisons will be made.

Prof. Wray Buntine joined Monash University in February 2014 after 7 years at NICTA in Canberra Australia.  He was previously of Helsinki Institute for Information Technology from 2002, and at NASA Ames Research Center, University of California, Berkeley, and Google. He is known for his theoretical and applied work in document and text analysis, data mining and machine learning, and probabilistic methods. He applies probabilistic and non-parametric methods to tasks such as text analysis.  In 2009 he was programme co-chair of ECML-PKDD in Bled, Slovenia, and was programme co-chair of ACML in Singapore in 2012.  He reviews for conferences such as ACML, ECIR, SIGIR, ECML-PKDD, ICML, NIPS, UAI, and KDD, and is on the editorial board of Data Mining and Knowledge Discovery.

Tuesday 8 July 2014

How people behave is really the central question for data analytics.  The way people play, the ways they interact, the kinds of behaviors they bring to the game ultimately drive how our systems perform, and what we can understand about why they do what they do.  In this talk I’ll describe three different scales of collecting data about user behavior, showing how looking at behavior data at the micro-, meso-, and macro-levels is a superb way to understand what people are doing in our systems, and why.  Knowing this lets you not just understand what’s going on, but also how to improve the user experience for the next design cycle. 

Daniel Russell is the Uber Tech Lead for Search Quality and User Happiness in Mountain View. He earned his PhD in computer science, specializing in artificial intelligence until he realized that magnifying and understanding human intelligence was his real passion. Twenty years ago he foreswore AI in favor of HI, and enjoys teaching, learning, running and music, preferably all in one day. He worked at Xerox PARC before it was PARC.com, and was in the Advanced Technology Group at Apple, where he wrote the first 100 web pages for www.Apple.com using SimpleText and a stone knife. He also worked at IBM and briefly at a startup that developed tablet computers before the iPad.

Wednesday 26 February 2014

This talk will present an overview of my recent research which evolves around discrete and computational differential geometry with applications in architecture, computational design and manufacturing. From the mathematical perspective, we are working on extensions of classical differential geometry to data and objects which frequently arise in applications, but do not satisfy the classical differentiability assumptions. On the practical side, our work aims at geometric modeling tools which include important aspects of function and fabrication already in the design phase. This interplay of theory and applications will be illustrated at hand of selected recent projects on the computational design of architectural freeform structures under manufacturing and structural constraints. In particular, we will address smooth skins from simple and repetitive elements, self-supporting structures, form-finding with polyhedral meshes, optimized support structures, shading systems and the exploration of the available design space.

Helmut Pottmann earned a Ph.D. in Mathematics from Vienna University of Technology in 1983. He has held faculty positions in Germany (Kaiserslautern, Hamburg) and the US (UC Davis, Purdue) and has been Professor of Applied Geometry at Vienna University of Technology since 1992. In 2009 he became Professor at King Abdullah University of Science and Technology, where he served as Director of the Geometric Modeling and Scientific Visualization Center until 2013. Pottmann has co-authored two books and more than 200 articles in scientific journals. He is also co-founder and scientific director of Evolute GmbH, a company which offers services and software to industries facing challenges related to complex geometry.

Wednesday 4 September 2013

The information contained across many data sets is often highly correlated. Such connections and correlations can arise because the data captured comes from the same or similar objects, or because of particular repetitions, symmetries or other relations and self-relations that the data sources satisfy. This is particularly true for data sets of a geometric character, such as GPS traces, images, videos, 3D scans, 3D models, etc. We argue that when extracting knowledge from the data in a given data set, we can do significantly better if we exploit the wider context provided by all the relationships between this data set and a "society" or "social network" of other related data sets. We discuss mathematical and algorithmic issues on how to represent and compute relationships or mappings between data sets at multiple levels of detail. We also show how to analyze and leverage networks of maps, small and large, between inter-related data. The network can act as a regularizer, allowing us to benefit from the "wisdom of the collection" in performing operations on individual data sets or in map inference between them.

This "functorial" view of data puts the spotlight on consistent, shared relations and maps as the key to understanding structure in data. It is a little different from the current dominant paradigm of extracting supervised or unsupervised feature sets, defining distance or similarity metrics, and doing regression or classification – though sparsity still plays an important role. The inspiration is more from ideas in homological algebra or algebraic topology,  exploiting the algebraic structure of data relationships or maps in an effort to disentangle dependencies and assign importance to the vast web of all possible relationships among multiple data sets. We illustrate these ideas largely using examples from the realm of 3D shapes and images -- but the notions are more generally to the analysis of graphs and other networks, acoustic data, biological data such as microarrays, homeworks in MOOCs, etc. This is an overview of joint work with multiple collaborators, as discussed in the talk.

Leonidas Guibas obtained his Ph.D. from Stanford under the supervision of Donald Knuth. His main subsequent employers were Xerox PARC, DEC/SRC, MIT, and Stanford. He is currently the Paul Pigott Professor of Computer Science (and by courtesy, Electrical Engineering) at Stanford University. He heads the Geometric Computation group and is part of the Graphics Laboratory, the AI Laboratory, the Bio-X Program, and the Institute for Computational and Mathematical Engineering. Professor Guibas' interests span geometric data analysis, computational geometry, geometric modeling, computer graphics, computer vision, robotics, ad hoc communication and sensor networks, and discrete algorithms. Some well-known past accomplishments include the analysis of double hashing, red-black trees, the quad-edge data structure, Voronoi-Delaunay algorithms, the Earth Mover's distance, Kinetic Data Structures (KDS),  Metropolis light transport, and the Heat-Kernel Signature. Professor Guibas is an ACM Fellow, an IEEE Fellow and winner of the ACM Allen Newell award.

Friday 18 January 2013

Limits in computing power and our ability to interact with computers have also imposed limits on our understanding of the world around us.  Increasingly, those limits are being removed, clearing the way for new advances in almost every kind of human endeavor.

Rick Rashid, Microsoft chief research officer and head of Microsoft Research, will present his vision of the future of computing research in light of these breakthroughs and the opportunities that lie ahead.

Tuesday 15 January 2013

It's very hard to understand the field of network protocols by focusing on the details of one particular protocol. Issues are clouded by marketing hype and protocol group rivalry. What is really intrinsic to the differences between one protocol and another? This talk covers some of the ways in which solutions can differ, as well as demystifying some especially confusing pieces of this field, such as what is really the difference between "layer 2 solutions" and "layer 3 solutions", why we need both Ethernet and IP, the evolution of Ethernet from its original invention (CSMA/CD) through spanning tree and now TRILL, and some things that people assume to be true that may not be. The talk includes some possible research areas.

Radia Perlman is a Fellow at Intel Labs, specializing in network protocols and security protocols.  Many of the technologies she designed have been deployed in the Internet for decades, including link state routing, the spanning tree algorithm, and TRILL, which improves upon spanning tree while still "being Ethernet".  She has also made contributions to network security, including assured delete of data, design of the authentication handshake of IPSec, trust models for PKI, and network infrastructure robust against malicious trusted components. She is the author of the textbook "Interconnections: Bridges, Routers, Switches, and Internetworking Protocols", and co-author of "Network Security". She has a PhD from MIT in computer science, holds over 100 issued patents, and has received various industry awards including lifetime achievement awards from ACM's SIGCOMM and Usenix, and an honorary doctorate from KTH.

Wednesday 3 October 2012

We all have a pet behaviour we would like to change, such as eating better, exercising more, or reducing our energy consumption. Many of us would also like to manage our time more effectively, by spending less time randomly Googling, sofa slouching or looking out the window. How can we design new technologies to help people change their behaviour? Nudging methods, derived from behavioural economics and social psychology, have become increasingly popular. But how effective are they and can technology be designed to exploit them? In this talk, Yvonne will describe our investigations into how decision environments can be restructured in innovative ways, using pervasive, ambient and wearable technologies to nudge behaviour in ways that are desirable to the individual. Our goal is to help people make better-informed decisions in situ. Underlying all of this, however, is the nagging question of whether it is ethical, desirable or sustainable to be nudging people in a desired direction. Or, is it a case of technological fudging, where we may be covering over deeper problems?

Yvonne's research interests are in the areas of ubiquitous computing, interaction design and human-computer interaction. A central theme is how to design interactive technologies that can enhance life by augmenting and extending everyday, learning and work activities. This involves informing, building and evaluating novel user experiences through creating and assembling a diversity of pervasive technologies. Yvonne has been awarded a prestigious EPSRC dream fellowship and is currently (until June 2012) rethinking the relationship between ageing, computing and creativity. Yvonne is also visiting Professor at the Open University, Indiana University, and Sussex University, and has spent sabbaticals at Stanford, Apple, Queensland University, and UCSD. Central to her work is a critical stance towards how visions, theories and frameworks shape the fields of HCI, cognitive science and Ubicomp. She has been instrumental in promulgating new theories (e.g. external cognition), alternative methodologies (e.g. in the wild studies) and far-reaching research agendas (e.g. "Being Human: HCI in 2020 manifesto).

Wednesday 14 September 2012

The principles of information processing in the brain are still far from understood. But progress in computer technology means that we can now realistically contemplate building computer models of the brain that can be used to probe these principles much more readily than is feasible, or ethical, with a living biological brain. What might these models tell us about brain function, and what might we learn that can then be applied to building more efficient, fault-tolerant, parallel computers?

Steve Furber CBE, FRS, FREng is the ICL Professor of Computer Engineering at the School of Computer Science at the University of Manchester and is probably best known for his work at Acorn Computers, where he was one of the designers of the BBC Micro and the ARM 32-bit RISC microprocessor.

Wednesday 21 March 2012

Download the slides here

Shari Lawrence Pfleeger is Director of Research for the Institute for Information Infrastructure Protection at Dartmouth College. She joined the I3P after serving for almost nine years as a senior researcher at the RAND Corporation. Previously, she headed Systems/Software, Inc., a consultancy specializing in software engineering and technology.  She has been a developer and maintainer for real-time, business-critical software systems, a principal scientist at MITRE Corporation's Software Engineering Center, and manager of the measurement program at the Contel Technology Center. She has also held several research and teaching positions at universities world-wide.

Shari is well-known for her work in empirical studies of software engineering and is the author of many books and articles, including Analyzing Computer Security (with Charles P. Pfleeger), Security in Computing (4e, with Charles P. Pfleeger), and Software Engineering: Theory and Practice (4e, with Joanne Atlee). She has been associate editor of IEEE Transactions on Software Engineering, associate editor-in-chief of IEEE Software, and she is currently associate editor-in-chief of IEEE Security & Privacy. Shari has been named repeatedly by the Journal of Systems and Software as one of the world's top software engineering researchers..  Shari earned a BA in mathematics from Harpur College, an MA in mathematics from Penn State, an MS in planning from Penn State, a PhD in information technology and engineering from George Mason University, and was awarded a Doctor of Humane Letters by Binghamton University.