----------------------------------------------------------------
CALL FOR PAPERS

NIPS 2012 Workshop: Machine Learning Approaches to Mobile Context Awareness
Lake Tahoe, Nevada, USA, December 8th 2012
https://sites.google.com/site/nips2012contextawareworkshop/
email: nips2012contextaware@gmail.com
----------------------------------------------------------------

Important Dates:

Submission Deadline: Friday, October 5th (extended)
Acceptance Notification: Wednesday, October 10th

---------------------------
Workshop Overview:

The ubiquity of mobile phones, packed with sensors such as
accelerometers, gyroscopes, light and proximity sensors, BlueTooth and
WiFi radios, GPS radios, microphones, etc., has brought increased
attention to the field of mobile context awareness. This field
examines problems relating to inferring some aspect of a user’s
behavior, such as their activity, mood, interruptibility, situation,
etc., using mobile sensors. There is a wide range of applications for
context-aware devices. In the healthcare industry, for example, such
devices could provide support for cognitively impaired people, provide
health-care professionals with simple ways of monitoring patient
activity levels during rehabilitation, and perform long-term health
and fitness monitoring. In the transportation industry they could be
used to predict and redirect traffic flow or to provide telematics for
auto-insurers. Context awareness in smartphones can aid in automating
functionality such as redirecting calls to voicemail when the user is
uninterruptible, automatically updating status on social networks,
etc.., and can be used to provide personalized recommendations.

Existing work in mobile context-awareness has predominantly come from
researchers in the human-computer interaction community. There the
focus has been on building custom sensor/hardware solutions to perform
social science experiments or solve application-specific problems. The
goal of this workshop is to bring the challenging inferential problems
of mobile context awareness to the attention of the machine learning
community. We believe these problems are fundamentally solvable. We
seek to get this community excited about these problems, encourage
collaboration between people with different backgrounds, explore how
to integrate research efforts, and discuss where future work needs to
be done. We are looking for participation both from individuals with
machine learning backgrounds who may or may not have attacked context
awareness problems before, and individuals with application-specific
backgrounds. Although the dominant mobile sensing platform these days
is the smartphone, we also welcome contributions that work with data
from a variety of body-worn sensors including standalone
accelerometers, GPS, microphones, EEG, ECG, etc., and custom hardware
platforms that combine multiple sensors. We are particularly
interested in contributions that deal with inferring context by fusing
information from different sensor sources.

In particular, we would like the workshop to address the following topics:

(1) What is the best way to combine heterogeneous data from multiple
sensors? Is contextual information encoded in specific correlation
patterns, or is there one sensor that “says it all” for each context,
and can we learn this automatically? How do we model and analyze
correlations between heterogeneous data?

(2) Feature extraction: what are the features that best characterize
these new sensor streams for analysis and learning? In video and
speech processing, such features have emerged over the years and are
now commonly accepted – are there certain features best suited for
accelerometer, audio environment, and GPS data streams? Can we learn
them automatically?

(3) A major part of this workshop will be dedicated to the discussion
of data. The community has a great need for a shared public dataset
that will allow researchers to compare algorithms and improve
collaboration. In our panel discussion we will discuss issues such as
creating a central data repository, common data collection apps, and
unique issues with context-awareness data.

---------------------------------
Submission instructions:

Papers may describe either novel or previously published/presented
work or works being presented at the main conference. Please use the
formatting guidelines for the NIPS conference. Submissions need not be
anonymous.

Selected papers will be assigned to either oral or poster presentations.

Please email your submissions to nips2012contextaware@gmail.com by
Friday, October 5th 23:59 PST.


---------------
Organizers:

- Katherine Ellis (UC San Diego)
- Gert Lanckriet (UC San Diego)
- Tommi Jaakola (MIT)
- Lenny Grokop (Qualcomm)