Close Enough: streaming algorithms for a world with limitations

Contributed Talk | Day 3 | 1:30 pm | 40 Minute Duration | Grand Gallery Overlook G

Close Enough: streaming algorithms for a world with limitations

Contributed Talk | Day 3 | 1:30 pm | 40 Minute Duration | Grand Gallery Overlook G

High-volume data analysis can quickly become computationally and practically expensive. Fortunately, we can benefit from decades of research in streaming algorithms: single-pass, approximate algorithms with bounded memory requirements and quick computation.

This talk introduces some of the major streaming approaches to:

  • approximate counting
  • the count-distinct problem
  • most frequent items
  • approximate percentiles

Though discipline is quite mathematical, this presentation will focus on intuitive explanations of the clever underlying logic behind these approaches. Attendees should also leave with some practical ideas about how approximate streaming algorithms can be applied.