Keynotes
Statistical Machine Learning Problems in the Brain Sciences
Robert E. Kass
Maurice Falk Professor of Statistics and Computational Neuroscience
Department of Statistics & Data Science and Machine Learning Department
Interim Co-Director, Center for the Neural Basis of Cognition
Carnegie Mellon University
Abstract: Machine learning may be considered the intersection of statistics and computer science. As a trained statistician who has participated actively in Carnegie Mellon's Machine Learning Department, I have been aware of the distinct perspectives these fields bring to machine learning, and would like to share some of my thoughts. I have also been involved in the brain sciences for roughly the past 20 years and will use several interesting examples drawn from neurophysiology to illustrate what I have to say. Finally, I will outline what I consider to be one of the most pressing and deep problems in the brain sciences, and will give reasons why good solutions will require new methods in statistical machine learning..
Towards an era of intelligent interactive algorithms
Dr. Aarti Singh
Associate Professor
A. Nico Habermann Faculty Chair 2013-2016
Machine Learning Department
Carnegie Mellon University
Abstract: Classical machine learning algorithms focus on the setting where the algorithm has access to a fixed dataset obtained prior to any analysis. In most applications, however, we have control over the data collection process such as which image labels to obtain, which drug-gene interactions to record, which network routes to probe, which movies to rate, etc. Furthermore, most applications face budget limitations on the amount and type of labels, data or features that can be collected. Decisions about which data to collect are typically taken by humans in an ad-hoc manner. Thus, there is a need to develop intelligent algorithms that can make principled and automated decisions to interact with the data generating mechanism and collect data that is most relevant for the learning task. In this talk, we ask the question - what does the freedom to interactively collect data buy us? I will present a sampling of work by my group on interactive methods for several learning problems such as regression, classification, matrix and tensor completion/approximation, column subset selection, learning structure of graphical models, reconstructing graph-structured signals, and clustering, as time permits. I will quantify the precise improvement in data efficiency, as well as demonstrate that interactive algorithms often also enable us to handle a larger class of data models. Finally, I will conclude with open directions and challenges that face interactive data analytics.