Important Dates

  • Extended submission deadline: Friday, November 4
  • Notification of acceptance: November 15
  • Workshop: Saturday, December 10, 7:30am - 10:30am & 4:00pm - 7pm (Day 2 of Workshops)

Organizers:

Overview of the Workshop

Recognition of high-level events from data streams is a much needed resource towards building intelligent machines that provide automatic and autonomous support in our every day lives. Recently, there has been significant progress towards recognition and discovery of human activities and interactions from data. Research progress is evident in (a) building real systems that extract information from a variety to sensors, to (b) developing and extending statistical machine learning approaches to model data for recognition and matching, to (c) study of specific domains to extract relevant higher-level context which in turn can be leveraged to support recognition. Many well-defined application domains have also been brought to the fore-front recently that only suggests the vast importance of automatic interpretation of data to recognize activities, actions, and behaviors of individuals and groups over both short and extended periods. These include surveillance and security, aides for older adults and children with special needs, automatic summary of soldier activities, long-term observation of cognitively disabled people, support of our every day lives, and tools for improving organizational efficiency and information flow. However, many significant questions in this area remain and require bringing together experts from various fields to come together to discuss upcoming challenges. Based on the successful NIPS-04 Workshop on Activity Recognition and Discovery, this workshop will again bring together experts from machine learning, sensing and perception, and ubiquitous computing to discuss issues related to

  1. What are the most recent developments in the field? What sensing technologies are available today and how can different forms of sensing technologies be brought to bear on this problem?
  2. What data modeling, data interpretation, and machine learning techniques are available today that can be applied to this problem? Which approaches proved successful, which ones did not work, and why?
  3. How can learning be facilitated and how do we deal with unknown and "surprising" events that were unexpected and therefore not modeled? How can we learn activity models in an unsupervised manner?
  4. What are the additional research questions in the context of understanding people's actions and interactions with other people? How can machine perception and machine learning support studies of social networks and social dynamics?
  5. What are some application domains that can leverage from this effort? What specific domain knowledge do these applications provide that can be used to focus the recognition task and perhaps aid in making it tractable?
  6. What should be the long-term and short-term goals of the research in modeling activities and interactions?

Workshop Format

The format of this workshop will be heavily biased towards discussions and brainstorming. We will have several presentations and a poster session.

Tentative list of invited speakers

  • Hung Bui, SRI International, Artificial Intelligence Center
  • Thomas Dietterich, Oregon State University, School of Electrical Engineering and Computer Science
  • David Hogg, University of Leeds, School of Computing
  • Andrew McCallum, University of Massachusetts Amherst, Department of Computer Science
  • Nuria Oliver, Microsoft Research, Adaptive Systems & Interaction group

Call for Papers

We seek novel research contributions in all areas related to activity recognition and discovery. Syntheses and survey papers are welcome, as long as they offer a fresh perspective on the field. Since there are only limited slots for oral presentations, please indicate in your submission whether you would be willing to present your work as a poster.
Submissions should follow the NIPS-05 paper format (8 pages). Poster abstracts should be 1-2 pages long. Authors should submit their work per email to nips-ard@cs.washington.edu.