The seminar this semester with the title "Selected Topics on the Hardware Implementation of Neural Networks" will focus on " Neuromorphic Architectures - A Feasibility Study for Real-Time Edge Computing".
The concept of "Neuromorphic Computing" describes a multi-disciplinary area that deals with the development of neural systems and targets to improve conventional methods from the field of machine learning. By adapting biologically inspired / realistic concepts, systems are derived, which will inherit the cognitive abilities and energy efficiency of their inspiration - the human brain. The seminar focuses on the state-of-the-art identifying state-of-the-art methods and evaluate / critically question the viability of those in the context of practicality and application contexts.
In this semester, a variety of neural systems will be examined for capability to process time-serial signals such as ECG data. Especially, with the rising popularity of "wearable devices" it needs to be clearified, if these algorithms and their specialized hardware components are suited for resource restrictive environments (e.g. in terms of energy, chip area etc.) and if there is optimization potential. Additionally topics in the area of security are offered.
The topical areas covered in this semester target diverse design principles and accelerators from the state-of-the-art, which are investigated in-depth and critically evaluated.
Further details and additional topics will be announced in the introduction session.
Registration is possible via RWTH Online.
For capacity reasons, the number of participants for this seminar is limited to 8.
Preliminary information on the seminar in SS 22 is held on 5/4/2022, 9:00 a.m. at seminar room UMIC 324 (2165|324)
Access data for the videoconference in case of online or hybrid mode will be announced in RWTH-Moodle or via e-mail. Participation in this preliminary meeting is mandatory.
During the semester, there will be 6 discussion sessions at 2-week intervals. The exact form and dates will be announced in RWTH-Moodle. Attendance at 5 out of 6 dates is mandatory.
For the certificate a written paper and a final presentation have to be given. The presentations will take place at 12/7/2022 from 9:00 am. Participation in all presentations is mandatory.