Monthly Archives: October 2017

Gao in CSSI seminar Fri. Oct. 13 12:30-2

The UMass Computational Social Science Institute invites you to an exciting talk by one of our own affiliates:

Song Gao
Associate Professor of Civil and Environmental Engineering, University of Massachusetts
Friday, October 13, 2017 • 12:30 p.m.-2:00 p.m. (lunch served at 12:15)
Computer Science Building, Room 150/151

Title: Travel Decision Making in an Uncertain, Dynamic, Information-Rich Urban Network

Abstract: Transportation systems are uncertain due to disruptions such as incidents and inclement weather. Real-time information, an integral element of smart cities, allows travelers to adapt to traffic conditions and potentially mitigate the adverse effects of uncertainty. Travelers create and suffer from congestion. The collective of all travelers’ choices defines the traffic load and its distribution. Understanding the choices made by travelers in an uncertain and dynamic environment who have access to real-time traffic information is paramount to better system design, management, and policy making. I have conducted a series of empirical studies using data from both laboratory experiments and in-vehicle tracking and monitoring devices in real-life urban networks. In this talk, I will focus on the modeling and optimization of adaptive route choice based on GPS (Global Positioning System) data from Stockholm, Sweden. It is found that two types of travelers both exist: one adapts to real-time information, and the other does not. Travelers are more likely to adapt during longer trips. Results from laboratory experiment suggest that travelers learn to adapt at a slower pace in a more complex network. An ongoing project on route optimization for taxi drivers’ customer searching using GPS data from Shanghai, China will also be discussed.

Bio: Song Gao is an associate professor of Civil and Environmental Engineering at the University of Massachusetts Amherst. Dr. Gao’s research focuses on transportation network optimization, econometric and psychological models of traveler behavior, equilibrium analysis of transportation systems with traveler information, with applications in smart and shared mobility, transportation planning under both normal and emergency conditions, and sustainable transportation systems. Prior to joining the faculty of the University of Massachusetts Amherst in 2007, Dr. Gao worked as a transportation engineer at Caliper Corporation, Newton, MA for three years, and developed advanced traffic assignment modules for TransCAD, a GIS-based transportation planning software. Dr. Gao received her Ph.D. and M.S. in Transportation from Massachusetts Institute of Technology in 2005 and 2002 respectively. She received her B.S. in Civil Engineering from Tsinghua University of China in 1999.

Jampani in MLFL Thurs. 10/12 at 11:45

Varun Jampani (Nvidia) will present “Bilateral Neural Networks for Image, Video and 3D Vision” in the Machine Learning and Friends Lunch Thursday Oct. 12 from 12:00pm to 1:00pm in CS 150. Abstract and bio follow.

Abstract:

Natural images exhibit high information correlation across pixels. Bilateral filtering provides a simple yet powerful framework for information propagation across pixels. The common use-case is to manually choose a parametric filter type, usually a Gaussian filter. We generalize the parameterization using a high-dimensional linear approximation and derive a gradient descent algorithm so the filter parameters can be learned from data. We demonstrate the use of learned bilateral filters in several applications where Gaussian bilateral filters are traditionally employed where we consistently observed improvements with filter learning. In addition, the ability to learn generic high-dimensional sparse filters allows us to stack several parallel and sequential filters like in convolutional neural networks (CNN) resulting in a new breed of neural networks which we call ‘Bilateral Neural Networks’ (BNN). We demonstrate the use of BNNs on several 2D, video and 3D vision tasks. Experiments on diverse datasets and tasks demonstrate the use BNNs for a range of vision problems.

Short Bio:

Varun Jampani works as a research scientist at Nvidia Research in Westford, US. He obtained his PhD at Max Planck Institute for Intelligent Systems (MPI) in Tübingen, Germany under the supervision of Prof. Peter V. Gehler. He works in the areas of machine learning and computer vision and his main research interests include probabilistic inference and neural networks. He obtained his BTech and MS from International Institute of Information Technology, Hyderabad (IIIT-H), India, where he was a gold medalist. During his studies, he did internships at Microsoft research institutes in Redmond (US), Cambridge (UK) and Cairo (Egypt); MPI, Tübingen (Germany) and; GE global research, Bangalore (India). He also worked as a volunteer teacher in Tibetan Children’s Village, Dharamsala, India.

Sullivan in CSSI seminar Fri. Oct. 6 12:30-2

The UMass Computational Social Science Institute invites you to an exciting talk by one of our own affiliates:

Florence Sullivan
Associate Professor of Education, University of Massachusetts
Friday, October 6, 2017 • 12:30 p.m.-2:00 p.m. (lunch served at 12:15)
Computer Science Building, Room 150/151

Title: Microgenetic Learning Analytics: Exploring Computational Means to Support Research on Collaborative Learning

Abstract: Microgenetic analysis is an educational research method that aims to capture and describe the development of conceptual understanding in children over short periods of time (minutes, hours, days). The method requires high density observations that span the period of rapidly changing competence. The level of explanatory power made possible by microgenetic analysis concerning the development of higher ordered thinking processes is unparalleled by other analytic techniques. However, the time-intensive nature of data collection and analysis using microgenetic techniques has, thus far, limited the generalizability of findings. To address this issue, we have begun to explore computational methods for supporting microgenetic analysis. Using the natural language processing approach of parts-of-speech tagging, we developed and implemented a new educational research method termed “Microgenetic Learning Analytics (MLA).” Results of research on student collaborative learning with educational robotics using the MLA method will be discussed.

Bio: Florence R. Sullivan is an associate professor of learning technology in the College of Education at UMass, Amherst. Dr. Sullivan received her Ph.D. in Cognitive Studies in Education from Teachers College, Columbia University. Her work focuses on student collaborative learning with computational media; with a special emphasis on girls’ learning and engagement with robotics. The aim of Dr. Sullivan’s research is to aid national efforts to improve the diversity of individual’s involved in computer science through educational interventions.

Orabona in MLFL Thurs. 10/5 at 11:45

Francesco Orabona (Stony Brook University) will present “Coin Betting For Backprop Without Learning Rates And More” in the Machine Learning and Friends Lunch Thursday Oct. 5 from 12:00pm to 1:00pm in CS 150. Abstract and bio follow.

Abstract:

Deep learning methods achieve state-of-the-art performance in many application scenarios. Yet, these methods require a significant amount of hyperparameters tuning in order to achieve the best results. In particular, tuning the learning rates in the stochastic optimization process is still one of the main bottlenecks. In this talk, I will propose a new stochastic gradient descent procedure that does not require any learning rate setting. Contrary to previous methods, we do not adapt the learning rates nor we make use of the assumed curvature of the objective function. Instead, we reduce the optimization process to a game of betting on a non-stochastic coin and we propose an optimal strategy based on a generalization of Kelly betting. Moreover, this reduction can be also used for other machine learning problems. Theoretical convergence is proven for convex and quasi-convex functions and empirical evidence shows the advantage of our algorithm over popular stochastic gradient algorithms.

Bio:

Francesco Orabona is an Assistant Professor at Stony Brook University. His research interests are in the area of theoretically motivated and efficient machine learning algorithms, with emphasis on online and stochastic methods. He received the PhD degree in Electrical Engineering at the University of Genoa, in 2007. He is (co)author of more than 60 peer reviewed papers.

Jordan in Cognitive Bag Lunch noon Weds. 10/4

Nancy Jordan (University of Delaware) will present “Living on the number line:  Development of numerical magnitude understanding in children at risk for learning difficulties in mathematics” in the Cognitive Bag Lunch series, Wednesday at noon in Tobin 423. All are welcome!