|
Machine Learning Internships Intel Research, Qualifications: Outstanding and highly motivated interns are needed to help develop a system for “user-activity based
adaptive power management”. Qualified candidates should be
familiar in general supervised and unsupervised machine learning. They should
also be familiar in learning and inference algorithms for graphical
models such as dynamic Bayesian networks and conditional random
fields. Finally, they should be knowledgeable with planning under
uncertainty algorithms, examples of which include Markov decision
processes (MDPs), partially observable MDPs and reinforcement
learning. Project goal: Adaptive or dynamic power management algorithms aim to reduce mobile platform (e.g. laptop) power consumption under performance constraints by selectively placing devices to
low power states. The success of such algorithms is critically
dependent on accurate modeling of the requests for service to the various
devices. We believe that such accurate modeling can only be achieved
by taking into consideration the user activity. Our first goal is to
explore learning algorithms for predicting user activity from sensor
readings (e.g. heyboard and mouse activity, applications running)as
well as predicting how the activities affect the requests for
service to the various devices. Our second goal is to explore algorithms
for computing power management policies given our predictive
models. Both goals may be challenging due to the inherent uncertainty
involved in prediction and the fact that some of the user activities may
not always be directly observable through sensors. All of our
leaning and control algorithms will be demonstrated and evaluated in
stages, on real mobile platforms. To apply please send your CV to [EMAIL PROTECTED] Thanks, Georgios |
_______________________________________________ uai mailing list uai@ENGR.ORST.EDU https://secure.engr.oregonstate.edu/mailman/listinfo/uai
