Streaming Algorithms BloomFilter

Probabilistic Data Structures for Summarization Instead of storing elements, streaming algorithms use mathematical properties and hashing to maintain compact summaries of the data. Bloom Filters (Membership Testing): When an algorithm needs to quickly check if an item has been seen before, it uses a Bloom filter. This structure uses a highly space-efficient bit array and multiple hash functions. When an item arrives, it is hashed, and corresponding bits are set to 1. It guarantees no false negatives (it will never miss an item actually in the set) but allows a small, tunable rate of false positives. Cuckoo Filters: Similar to Bloom filters but inspired by Cuckoo Hashing, these store compact representations of elements called "fingerprints" in buckets. If a bucket is full, an existing fingerprint is randomly evicted ("cuckooed") to an alternate location. This allows Cuckoo filters to uniquely support the deletion of elements, which standard Bloom filters cannot do. Count-Min Sketch (Frequency Estimation): To count how often items appear without storing a massive frequency table, this algorithm uses a sub-linear 2D array of counters. Each incoming item is passed through multiple hash functions, and the corresponding counters are incremented. To query an item's frequency, the algorithm returns the minimum value among its counters, providing an overestimate that becomes more accurate as more memory is allocated.