Guided Research - Project Proposal

Echo State Networks (ESN) have been used successfully in a broad range of applications, from dynamic pattern recognition to channel equalization in digital communication systems [5, 7, 9]. Their simplicity and ease of use, paired with their underlying ”mathematical power” make them an ideal choice in many black-box modelling tasks. For many applications, however, it is mandatory to learn and adjust the ESN parameters ”online”, i.e. during the actual task that the ESN is supposed to perform. A good example is high-frequency digital communications with cellular phones, where an ESN could be used to cancel out the noise that is injected into the radiofrequency signal. As the characteristics of the noise changes over time (since sender and receiver are moving), the ESN needs to change its internal parameters adaptively. So far, certain properties of ESNs make online learning difficult and slow. The proposed project aims at adjusting ESNs such that they become suitable for standard online learning methods by changing parts of the ESN’s internal structure. If we succeed in our efforts, this could for example lead to cell phone systems that are much more noise-resistant, therefore leading to clearer communications and higher data rates. Another envisioned area of application is the adaptive control and identification of highly nonlinear systems. This could, for example, be used to help a robot navigate autonomously in an unknown environment.