
Maria Illing
Referentin ZMN
+ 49 3677 69 3400
Gustav-Kirchhoffstraße 7
98693 Ilmenau
Feynmanbau (ZMN),
Raum 304
TU IlmenauThe Ilmenau School of Green Electronics (ISGE), a central research and training structure at TU Ilmenau, held its first workshop. Dr. Lina Jaurigue, one of the PIs of ISGE and Member of the Group for Theoretical Physics II, successfully organised a two-day workshop series in the Center of Micro- and Nanotechnologies (ZMN) on the topic of Reservoir Computing at the beginning of October.
The ISGE is dedicated to the development of sustainable, climate-neutral microelectronics and promotes interdisciplinary projects that reduce resource and energy consumption in the IT sector. Young scientists are researching innovative technologies that take the entire material cycle into account - a crucial component on the way to a sustainable future.
On 7 and 8 October, the focus was on the concept of reservoir computing. This is a neural network model that impresses with its simple structure and yet high efficiency: Only the output layer is trained, while the internal dynamics of a randomly constructed network are used to recognise complex patterns and make predictions.
Workshop highlights:
📅 Monday, 07.10.2024
9:00 - 10:30: Introduction to Reservoir Computing. Here the basics of how reservoirs are trained and used for prediction or classification were explained. There were also interesting insights into related technologies such as Extreme Learning Machines (ELM) and non-linear vector regression.
10:45 - 12:00: Practical work with the Jupyter Notebook. Participants created a discrete-time reservoir and worked step-by-step with input and target arrays, training and test state matrices as well as output weights and predictions.
13:00 - 14:00: Discussion on the influence of task implementation and how input data influences the predictions of the system.
📅 Tuesday, 08.10.2024
9:00 - 9:30: Dynamic properties of reservoirs - a comparison between discrete and continuous reservoirs.
9:30 - 10:30: Time multiplexing and delay-based reservoirs. Participants explored input methods and network analysis of delay-based reservoirs, with exciting parallels to spatially-multiplexed systems.
10:45 - 11:30: Overview of physical reservoir computing - an emerging field with great potential, but also challenges.
11:30 - 12:00: Concluding discussion. What kind of tasks are suitable for reservoir computing? How do the tasks influence the design and operation of a reservoir? The role of delay embedding was also discussed in depth.
Special thanks goes to the Carl Zeiss Foundation, whose generous support made it possible to realise this pioneering project. This workshop emphasise ISGE's commitment to supporting the next generations of researchers in the field of sustainable and future-proof solutions for the electronics industry.