About
Learn about the V4-ICARUS project
Learn about the V4-ICARUS project
V4-ICARUS is a scientific research project that intends to develop the next generation of mathematical signal processing and data analysis methods for the new era of big data. The innovative paradigm of V4-ICARUS will move well beyond the state-of-the art in both signal processing and machine learning, wherein instrument-specific methods are considered for monolithic tasks, such as classification and clustering. Within our project, modern signal processing (e.g. sparse representations, tensor completion), machine learning (e.g. deep learning), and distributed computing (e.g. in-memory analytics) techniques will be intertwined and evolve to encompass the complex Big Sensor Data characteristics (e.g. non-linearity, high-dimensionality, noisy artifacts, spatio-temporal evolving behaviors). The envisaged models will yield a mathematically rigid framework for selectively combining data from diverse, and possibly inhomogeneous, data streams.
Design efficient high-order tensorial structures for sparse encoding of high-dimensional multi-instrument Big Data sets.
Develop a novel algorithmic framework for uncertainty-aware processing of Big Sensor Data, recorded in decentralized, multi-instrument, multi-source environments.
Design a new generation of online distributed optimization techniques enabling Deep Signal Learning algorithms to handle high-velocity high-dimensional streams of complex heterogeneous Big Sensor Data.
Implement an innovative, highly reconfigurable, two-layer architecture (intra-cluster for optimal utilization of computational resources and inter-cluster for online execution of the data-driven algorithms) that will facilitate the execution of the algorithmic tools developed in Objectives 1-3.