FWF - SISE-NDML - Signal and Information Processing in Science and Engineering - Nonlinear Dynamic and Machine Learning

Distributed signals and data will be of great importance to our future daily life. The application of ubiquitous networked sensors, processing units, and distributed data sets will enhance the understanding of our world and its sustainable use. In this context, massive amounts of data have to be turned into concise and useful information, which demands groundbreaking new science at the intersection of mathematics, signal and information processing, communications theory, and scientific computing. We aims at developing new theories, algorithms, and implementations that enable the extraction, compression, transmission, and storage of information in large-scale distributed data sets. The focus will be on distributed architectures which can be designed to be fault tolerant and scalable. The concepts and methods that will result from this basic research will be applicable to sensor and communication networks, distributed systems, cooperative wireless communications, machine learning, embedded system design, medicine, and molecular biology. The research to be conducted in the proposed NFN SISE covers the following areas: - Flexible frame-based signal representations for distributed, nonstationary, and stochastic environments - Methods for distributed detection that are robust to signal compression and transmission errors - Nonlinear dynamics for sequential data modeling and spatio-temporal data fusion - Distributed joint source-channel-network coding with a minimum amount of node cooperation; - Cooperative communication systems for mobile users and sensors - Optimized hardware-adaptive numerical algorithms trading accuracy for parallel performance. - Design of the network and its nodes using a formalized mathematically tractable language.