July 12th 2016 (Tue.)15:30~16:30How Deep Learning Could Help 5G: Some Preliminary Ideas

Spearker:Professor Kung, Hsiang-Tsung
Topic:How Deep Learning Could Help 5G: Some Preliminary Ideas
Time:15:30 pm~16:30 pm July 12th 2016 (Tue.)
Venue:Room 528, College of Engineering IV, NCTU (up to 50 people)
Organizers:Big Data Research Center
Co-Organizers:College of Computer Science, College of Electrical and Computer Engineering, Microelectronics and Information Systems Research Center (MIRC)
The 5G standard for mobile networking will be a far greater departure from
 its predecessor, 4G, than 4G was of its predecessor. We expect a
 proliferation of small, overlapping, dense cells as well as use of
 millimeter wave (mmW) bands. Controlling such 5G systems will be
 unprecedentedly complex, which will likely defy conventional physical
 model-based solutions. Instead, data-driven modeling must be investigated.
 Can deep learning, which has enjoyed great success recently in other
 fields, play a role in 5G? In this talk, I will discuss some possibilities
 from my own group at Harvard, such as applying deep neural networks to (1)
 find a direction for reaching a target receiver over mmW and (2) select an
 optimal mmW beam in the environment with multiple antennas. In addition, I
 will describe recent results on speeding up feedforward inference over deep
 neural networks.