S3 - Asal Kiazadeh, Flexible AI hardware towards energy-efficient edge device for medical wearables
19 Outubro 2022, 14:00 • Maria Margarida Cruz
Abstract: Exponential increase of data volume due to smart
devices in the concept of internet-of-things (IoT) give rise to an
ultra-high energy demand. It is notable that the dominating factor of
power consumption is in fact related to data transfer between processing
unit and memory rather than computation itself. Therefore, solutions to
reduce power consumption to a large extent must rely on memory-centric
computation and brain-inspired neuromorphic architectures.
In
this seminar, flexible IC design for medical applications is targeted.
We showcase different synapse devices using amorphous oxide
semiconductors (AOS) technology, besides a proof-of-concept of on-chip
matrix vector multiplication, applying memristor technology.