HyperLogLog (Cardinality Estimation): This algorithm estimates the number of distinct elements in a stream. It hashes items into binary strings and looks for the position of the leftmost 1-bit (a pattern indicating how rare an item is). To reduce variance, it employs stochastic averaging, splitting the stream into substreams and averaging their results using a harmonic mean. This allows HyperLogLog to estimate cardinalities of over 1 billion with only 1.5 kilobytes of memory and an error rate of roughly 2%.
