Physical Reservoir Computing: AI in a bucket of water
Manish Yadav, Technische Universität Berlin, Fakultät V - Verkehrs- und Maschinensysteme
The project of Physical Reservoir Computing (RC) presents a groundbreaking and unconventional approach to constructing an energy-efficient machine learning method utilizing a physical system, specifically employing a bucket of water. Collaborative efforts among students will encompass machine learning, micro-controller programming, computer vision, and data processing, with the ultimate goal of creating a demonstrator for eco-friendly AI that capitalizes on the intricacies found in nature. Target group: A broad horizon of different backgrounds and expertise is required for the successful implementation of the project, as it involves deep interdisciplinary research. Degree courses across all disciplines of the natural sciences: mathematics, physics, computer science, engineering, informatics; and across Bachelor and Master level. Students need to have a prior knowledge of computer programming (Python, MATLAB), basic linear algebra and understanding of ML algorithms. A subset of students should be interested in electronics, micro-controller programming and hands-on experimental work. Why RC? RC has been successfully applied to many computational problems, such as temporal pattern classification, time series forecasting, pattern generation, adaptive filtering and control, and system identification, at the same time providing an environmentally friendly alternative to classical deep learning techniques, holding immense potential for eco-friendly AI.