Speaker: Prof. Neil Bergmann, University of Queensland, Australia

Title: Machine Learning and Sensor Networks

Abstract: Wireless sensor networks are becoming the eyes and ears (and other senses) of the Internet, allowing high temporal and spatial sampling of data from both the natural and the built environment.  The benefits of wireless operation often mean that such sensor nodes are battery powered, perhaps with some energy harvesting.  Usually, such sensors are limited in their temporal resolution by their limited energy.  "Dumb" sensors simply record and transmit raw transducer data streams for subsequent data analysis by powerful processors.  The majority of the energy used by such sensors is in the radio transmission of the raw data.  Communications energy can be saved, if the data can be compressed or otherwise on a "smart" sensor node, and only compressed or summary information sent, but this requires energy-efficient on-node processing.  This seminar summarises results from a past project using machine-learning techniques for on-sensor processing, and discusses proposals for how this on-sensor processing can be done in a more energy efficient fashion using reconfigurable hardware (FPGAs).

Biography: Prof. Neil Bergmann has been the Chair of Embedded Systems in the School of Information Technology and Electrical Engineering at the University of Queensland, Brisbane, Australia since 2001.  He has Bachelors degrees in Electrical Engineering, Computer Science, and Arts from the University of Queensland, and a PhD in Computer Science from the University of Edinburgh in 1984.  His research interests and in computer systems, especially reconfigurable computing and Wireless Sensor Networks.  He is on sabbatical leave, visiting University of Edinburgh during August 2016.