Frequency Estimation in Streams
73 words
1 min read
Visual companion
Python
Type and operator map
Python Week 1: the first filter for runtime behavior
View
Revision summary
What this note is really saying
Short form
# Frequency Estimation in Streams ## Count-Min Sketch vs Count Sketch Property Count-Min Count Estimate bias Overestimates (never underestimates) Unbiased Space $O(\frac{1}{\epsilon} \log \frac{1}{\delta})$ $O(\frac{1}{\epsilon^2} \log \frac{1}{\delta})$ Update time $O(\log \frac{1}{\delta})$ $O(\log \frac{1}{\delta...

Frequency Estimation in Streams
Count-Min Sketch vs Count Sketch
| Property | Count-Min | Count |
|---|---|---|
| Estimate bias | Overestimates (never underestimates) | Unbiased |
| Space | O(ϵ1logδ1) | O(ϵ21logδ1) |
| Update time | O(logδ1) | O(logδ1) |
| Error guarantee | $ |
Count-Min Sketch: Guarantee
For any query i, with probability 1−δ:
Where ∣∣f∣∣1 is the sum of all frequencies, ϵ is error parameter, δ is failure probability.
Join Discord
PreviousMatrix SketchingNextStreaming Algorithms