Very Fast Streaming Submodular Function Maximization

Data summarization has become a valuable tool in understanding even terabytes of data. Due to their compelling theoretical properties, submodular functions have been the focus algorithms. Submodular function maximization is well-studied problem with variety algorithms available. These usually offer worst-case guarantees expense higher computation and memory requirements. However, many practical applications do not fall under this mathematical but are much more well-behaved. We propose new algorithm called ThreeSieves that ignores thus uses fewer resources. Our selects most informative items from data-stream on fly maintains provable performance cases fixed budget. In an extensive evaluation, we compare our method against 6 state-of-the-art 8 different datasets including data without concept drift. show outperforms current majority and, at same time,