Znanost razvoja i korištenja, praćenja i korištenja elektroničkih instrumenata u svrhu mjerenja, praćenja i zabilježavanja različitih fizikalnih fenomena koji mogu, a i ne moraju biti električke prirode. To uključuje analogne i digitalne elektroničke instrumente, sustave i standarde za mjerenje i bilježenje električkih veličina u frekvencijskoj i vremenskoj domeni, te mjerne pretvornike neelektričkih veličina. Predmet zanimanja su također i instrumenti s automatiziranim upravljanjem i analitičkim funkcijama.
Odjel za instrumentaciju i mjerenja
Zavod za Elektroničke sustave i obradbu informacija, Hrvatska sekcija IEEE, Odjel za instrumentaciju i mjerenja i ZCI-DATACROSS, pozivaju Vas na predavanje
"FANNCortexM: An Open Source Library for Multi-layer Neural Network on ARM Cortex M family for low power smart sensing"
koje će održati Dr Michele Magno, ETH, Zurich, Švicarska
u četvrtak, 18.10.2018, u 13:40-14:25, u Sivoj vijećnici Fakulteta elektrotehnike i računarstva Sveučilišta u Zagrebu Unska 3, Zagreb.
Predavanje će se održati u okviru Međunarodne radionice o naprednim kooperativnim sustavima (ZCI-DATACROSS, IWACS 2018). Program radionice dsostupan je na: https://sites.google.com/view/iwacs2018/program
Više o predavanju i o predavaču pročitajte u opširnijem sadržaju obavijesti.
Sažetak predavanja:
Smart sensing is a promising technology to enhance user experience that has already been exploited in sport/fitness, as well as health and human monitoring. Smart sensing systems not only provide continuous data monitoring and acquisition, but are also expected to process, and make sense of the acquired data by classification in similar ways as human experts do. However often smart sensing devices are supplied by battery to be unobtrusive. Supporting continuous operation on ultra-small batteries poses unique challenges in energy efficiency. In this talk, I will present the Fast Artificial Neural Network (FANN) library and the porting and optimization done for ARM-Cortex M family. I will present also the energy and accuracy performance on an MSP432 and Ambiq Apollo 2 in different applications such as stress detection and emotion detection. Moreover, we evaluate the max size of the neural network on the two microcontrollers according to the on-chip memory. We demonstrate with experimental results that is possible to achieve up to 100% accuracy consuming few uJ. Finally, an example of the implementation on a smart bracelet that uses the Ambiq Apollo 2 and can be self-sustaining using energy harvesting will be shown. Our group will realize a open-source implementation of FANN for Arm-Cortex-M.
O predavaču:
Michele Magno (IEEE SM’16) received the masters’ and Ph.D. degrees in electronic engineering from the University of Bologna, Italy, in 2004 and 2010, respectively. He is currently a Senior Researcher with ETH Zürich, Switzerland at the Integrated System Laboratory and a Research Fellow with the University of Bologna, Italy. He has authored more than 140 papers in international journals and conferences. His current research interests include smart sensing, low power machine learning, wireless sensor networks, wearable devices, energy harvesting, low power management techniques, and extension of lifetime of batteries-operating devices.
