Final Report Summary - PLEASE (PLEASE: Projections, Learning, and Sparsity for Efficient data-processing)
A flagship application of sparsity is the paradigm of compressed sensing, which exploits sparsity for data acquisition using limited resources (e.g. fewer/less expensive sensors, limited energy consumption, etc.). Besides sparsity, a key pillar of compressed sensing is the use of random low-dimensional projections.
Sparse models and random projections are at the heart of many success stories in signal processing and machine learning. A major outcome of PLEASE is the demonstration of their high potential for resource-efficient learning at the era of Big Data, through the new concept of Compressive Learning. A particular example is Compressive Clustering, which makes it possible to cluster large collections (of typically tens of gigabytes or more) of training examples using a single concise ”summary” –called sketch– which size (typically only a few kilobytes) is independent of the number of examples, offering substantial memory savings.
With applications from audio processing and enhancement to brain imaging, PLEASE establishes a unifying framework to handle information from the sensing level to the semantic level through a series of mathematical, algorithmic, and experimental contributions that are deeply rooted in applied mathematics, statistics, and computer science.