Select a calendar:
Filter March Events by Event Type:
SUNMONTUEWEDTHUFRISAT
Events for March 03, 2015
-
CS Colloquium: Sergey Levine (UC Berkeley) - Deep Learning for Decision Making and Control
Tue, Mar 03, 2015 @ 09:45 AM - 10:50 AM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Sergey Levine, UC Berkeley
Talk Title: Deep Learning for Decision Making and Control
Series: CS Colloquium
Abstract: A remarkable feature of human and animal intelligence is the ability to autonomously acquire new behaviors. My work is concerned with designing algorithms that aim to bring this ability to robots and simulated characters. A central challenge in this field is to learn behaviors with representations that are sufficiently general and expressive to handle the wide range of motion skills that are necessary for real-world applications, such as general-purpose household robots. These representations must also be able to operate on raw, high-dimensional inputs and outputs, such as camera images, joint torques, and muscle activations. I will describe a class of guided policy search algorithms that tackle this challenge by transforming the task of learning control policies into a supervised learning problem, with supervision provided by simple, efficient trajectory-centric methods. I will show how this approach can be applied to a wide range of tasks, from locomotion and push recovery to robotic manipulation. I will also present new results on using deep convolutional neural networks to directly learn policies that combine visual perception and control, learning the entire mapping from rich visual stimuli to motor torques on a real robot. I will conclude by discussing future directions in deep sensorimotor learning and how advances in this emerging field can be applied to a range of other areas.
The lecture will be streamed through the dedicated link HERE.
Biography: Sergey Levine is a postdoctoral researcher working with Professor Pieter Abbeel at UC Berkeley. He completed his PhD in 2014 with Vladlen Koltun at Stanford University. His research focuses on robotics, machine learning, and computer graphics. In his PhD thesis, he developed a novel guided policy search algorithm for learning rich, expressive locomotion policies. In later work, this method enabled learning a range of robotic manipulation tasks, as well as end-to-end training of policies for perception and control. He has also developed algorithms for learning from demonstration, inverse reinforcement learning, and data-driven character animation.
Host: Computer Science Department
More Info: https://bluejeans.com/658994068
Location: Olin Hall of Engineering (OHE) - 132
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
Event Link: https://bluejeans.com/658994068
-
Epstein ISE Department Seminar
Tue, Mar 03, 2015 @ 10:00 AM - 11:00 AM
Daniel J. Epstein Department of Industrial and Systems Engineering
Conferences, Lectures, & Seminars
Speaker: Phebe Vayanos, Lecturer and Postdoctoral Associate, MIT Operations Research Center
Talk Title: Data-Driven Learning Under Uncertainty: An Adaptive Optimization Perspective
Abstract: Motivated by the recent explosion of data availability and the plethora of decision problems affected by uncertainty, we propose a data-driven paradigm for dynamic learning that unifies optimization and estimation. Our framework naturally captures the critical exploration-exploitation trade-off of the decision-maker, and we develop a tractable solution scheme to compute near-optimal policies. We showcase the versatility of our method by applying it to two very diverse areas: we focus on a pricing problem arising in revenue management and then discuss an application in energy.
In the area of revenue management, we discuss the pricing problem faced by a retailer who has a finite inventory of a product available for sale. We assume that the product demand curve is unknown to the retailer who has at his disposal a history of sales data. We present computational results that show that our proposed policies: (a) yield higher profits compared to commonly used policies, (b) nearly match results obtained with perfect information under downside measures such as Conditional Value-at-Risk, and (c) can be obtained in modest computational time for large-scale problems.
In the area of energy, we discuss an industrial application of our research in collaboration with BP, one of the worldâs major oil and gas companies. Using actual data from a BP oilfield, we create a simple and powerful model for predicting oil production that circumvents the need for complex reservoir modeling. We leverage this model and the framework described above to devise a methodology that enables oil companies to maximize the quantities of oil extracted from each reservoir, and therefore decrease the natural resources (and energy supplies) that are left untapped.
This is joint work with Dimitris Bertsimas, MIT.
Biography: Phebe Vayanos is a lecturer in the Operations Research and Statistics Group at MIT Sloan School of Management, and a postdoctoral research associate in the Operations Research Center at MIT. Her current research is focused on developing data-driven models and scalable solution approaches for real-world decision problems affected by uncertainty and ambiguity. In particular, she is motivated by applications in revenue management, energy, finance, education, and healthcare. She holds a PhD degree in Operations Research and an MEng degree in Electrical & Electronic Engineering, both from Imperial College London. She has extensive experience with the energy and investment banking industries, having worked at JPMorgan and BNP Paribas and having consulted for BP.
More Information: SEMINAR-Vayanos.doc
Location: Ethel Percy Andrus Gerontology Center (GER) - 206
Audiences: Everyone Is Invited
Contact: Georgia Lum
-
Epstein Institute / ISE 651 Seminar Series
Tue, Mar 03, 2015 @ 03:30 PM - 04:50 PM
Daniel J. Epstein Department of Industrial and Systems Engineering
Conferences, Lectures, & Seminars
Speaker: Alejandro Toriello, Assistant Professor, H. Milton Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology
Talk Title: The One-Dimensional Dynamic Dispatch Waves Problem
Series: Epstein Institute Seminar Series
Abstract: We study same-day delivery distribution systems by formulating the Dynamic Dispatch Waves Problem (DDWP), which models a depot where delivery requests arrive dynamically throughout a service day. At any dispatch epoch (wave), the information available to the decision maker is (1) a set of known, open requests which remain unfulfilled, and (2) a set of potential requests that may arrive later in the service day; the decision maker decides whether or not to dispatch a vehicle at each wave, and if so, which subset of open requests to serve, with the objective of minimizing expected vehicle operating costs and penalties for unserved requests. We consider the DDWP with a single delivery vehicle and request destinations on a line: We describe a class of a priori dispatch policies that plan routes for each wave in advance, and provide a dynamic programming approach for determining an optimal policy of this kind. We then discuss the benefits of dynamic policies, and propose several bounds and heuristics for the dynamic case.
Joint work with Alan Erera and Mathias Klapp
Biography: Alejandro Toriello joined Georgia Tech ISyE in August 2013 as an assistant professor. His research interests lie in the theory and application of supply chain management, logistics and transportation, and in related optimization methodologies. He currently serves as associate editor for the journals Optimization Methods and Software and Transportation Science. Prior to joining ISyE, he served as an assistant professor in the Epstein Department of Industrial and Systems Engineering at the University of Southern California.
Host: Daniel J. Epstein Department of Industrial and Systems Engineering
More Information: Seminar-Toriello2.docx
Location: Ethel Percy Andrus Gerontology Center (GER) - 206
Audiences: Everyone Is Invited
Contact: Georgia Lum
-
In & Out: 30 Minutes to Identify Internships & Jobs Still Available!
Tue, Mar 03, 2015 @ 05:00 PM - 06:00 PM
Viterbi School of Engineering Career Connections
Workshops & Infosessions
Looking for a job after graduation or an internship this summer? Join VCS for 30 minutes to learn about resources you can use to identify and apply for employment opportunities.
Location: Ronald Tutor Hall of Engineering (RTH) - 109
Audiences: All Viterbi Students
Contact: RTH 218 Viterbi Career Services
-
eBay Info Session
Tue, Mar 03, 2015 @ 05:30 PM - 07:00 PM
Viterbi School of Engineering Career Connections
Workshops & Infosessions
Join representatives of this company as they share general company information and available opportunities.
Location: Seeley G. Mudd Building (SGM) - 101
Audiences: Everyone Is Invited
Contact: RTH 218 Viterbi Career Services
