H2020 MSCA Innovative Training Network for the research on Advanced Hardware/Software Components for Integrated/Embedded Vision Systems


Efficient implementation of deep learning inference in order to bring Al to the edge

@ Instituto de Microelectrónica de Sevilla (CSIC-Universidad de Sevilla), Spain

In general, deep neural networks (DNNs) use simple mathematical operations, but the volume of these calculations is very large. In the inference level, where the neural network is trained and ready to use, the optimal allocation of hardware resources to perform the specific task is very complex process. Storing several weights requires more space than the internal memory capacity of current microprocessors (CPUs and GPUs). As a result, we have no choice but to store neural network weights in an out of processor memory. The primary goal of this project is bringing memory and computing processor closer to each other. This can be used for providing better performance in terms of energy efficiency and power consumption.

Hossein Khosravi



Hossein Khosravi was born in Mashhad, Iran on July 15, 1992. Hossein completed his Bachelor’s degree at University of Birjand  in the field of Telecommunication Engineering (2010-2014). He obtained his Master degree from University of Birjand  in Electrical Engineering (Analog Integrated Circuit design) (2016-2018).Since 2019 to 2020 he served as research assistant at Universiy of Minho, Portugal on the design of communication components for WLAN application such as Low noise amplifiers (LNAs) and Mixers. He is currently an early stage researcher and PhD student at Instituto de Microelectronica de Sevilla (IMSE-CNM), Sevilla, Spain. He is interested in Analog Integrated circuit Design, Microelectronics, image processing and ASIC.










"Efficient implementation of deep learning inference in order to bring Al to the edge"