The classic framework of machine learning is: example in, prediction out. This is great when examples are fully available at all times, and when all parts of an example are relevant for making a prediction. But it is very different from how humans reason. We get some information. We may make a prediction. Or we may decide we need to get more information.
Abstract: Observational healthcare data, such as administrative claims and electronic health records, play an increasingly prominent role in healthcare. Pharmacoepidemiologic studies in particular routinely estimate temporal associations between medical product exposure and subsequent health outcomes of interest and such studies influence prescribing patterns and healthcare policy more generally.
Speaker: Eleni Drinea - Adjunct Professor, Columbia University
Speaker: Adam Kapelner - PhD Candidate in Statistics, Wharton School of the University of Pennsylvania
Title: Matching on-the-fly: Sequential Allocation with Higher Power & Efficiency
Speaker: David Pritchard
Title: Teaching the Craft of Code
Abstract: Programming is a skill that is best learned actively. I will discuss several websites that I have developed with this aim: Computer Science Circles (Python), Websheets (Java), and a Java execution visualizer.
Abstract: The capacity of a communication channel is the maximum rate at which information can be reliably transmitted over the channel. In this work I consider the capacity of the binary deletion channel, where bits are deleted independently with a certain probability. This represents perhaps the simplest channel with synchronization errors but a characterization of its capacity remains an open question. I will present several techniques to lower bound the capacity, including Markov chain methods, Poisson-repeat channels, and ideas from renewal theory.
Abstract: In this talk, I will present a few urban data sets and discuss extracted patterns and models, and their variability. Based on vehicle routes of several data sets, I will then present a novel transport planning algorithm with transfers and show the corresponding resource savings. This work is in collaboration with students, in particular with Brian Coltin.
Biography: Manuela Veloso’s long-term research goal is the effective construction of autonomous agents where cognition, perception, and action are combined to address planning, execution, and learning tasks. Her vision is that multiple intelligent robots with different sets of complementary capabilities will provide a seamless synergy of intelligence. Manuela Veloso’s research focuses on the continuous integration of reactive, deliberative planning, and control learning for teams of multiple agents acting in adversarial, dynamic, and uncertain environments. Her multiagent and multirobot research interests have been motivated by and experimented in the domain of robot soccer. Since 2009, she has been investigating indoor mobile, service, companion robots, CoBots, such that robots and humans interact in a symbiotic relationship building upon individual strengths and limitations. Veloso created and directs her CORAL overarching research lab for the research on intelligent agents that Collaborate, Observe, Reason, Act, and Learn. As of 2010, she has ten PhD students and has graduated other twenty one PhD students, whose theses are available at her website www.cs.cmu.edu/~mmv. She thanks her students for the compelling research that they jointly pursue.
Speaker: Joachim M. Buhmann, Computer Science Department, Machine Learning Laboratory, ETH Zurich
Title: What is the Information Content of an Algorithm?