Summarizing Edge-Device Data via Core Items.

In & para;this work, we consider the problem of summarizing a data stream through an item-based summary using core items. We consider an IoT setting, where computing such summaries at the edge devices instead of emitting the whole data stream can drastically reduce the network traffic and speed up further processing. Core items of a data stream are the items with the highest values for a given monotone submodular utility function. To create stream summaries, we propose the SoftSieving approach for parallel processing with low memory consumption and fast execution time while attaining acceptable utility gain. Through extensive experiments with real-world datasets, we show the suitability of our approach and its superiority over state-of-the-art competitors.