heavy-hitters

[NeurIPS'23] H2O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models.

Probabilistic data structures in python http://pyprobables.readthedocs.io/en/latest/index.html

Reinforcement Learning (RL)-based routing algorithm for SDN networks created from scratch using Python.

Heavy-Hitter detection in P4 switch ASIC using Inter-Packet Gap

Sliding-window and regular top-K sketches, based on HeavyKeeper

Implementation for - Mitigating DNS random subdomain DDoS attacks by distinct heavy hitters sketches

Mastic: Private Weighted Heavy-Hitters and Attribute-Based Metrics

A novel approach based on Random Forests that quickly and accurately identifies elephant flows in programmable Data Planes.

Cuckoo Heavy Keeper and the balancing act of maintaining heavy hitters in stream processing

Code for "PLASMA: Private, Lightweight Aggregated Statistics against Malicious Adversaries"

Source code of "Network-Wide Routing-Oblivious Heavy Hitters" paper by Ran Ben Basat, Gil Einziger, Shir Landau Feibish, Jalil Moraney, and Danny Raz (ACM/IEEE ANCS 2018).

Demo app for the https://github.com/keilerkonzept/topk sliding top-k sketch library

Efficient Flow Recording with InheritSketch on Programmable Switches (ICDCS2023)

Count distinct LLM users and find top token consumers across services, days and tools. Fast, small, portable summaries in Go and Python.

Implementation of streaming algorithms (Misra-Gries & Lossy Counting) for getting frequent items from data streams.

Probabilistic data structures for JavaScript/TypeScript: blocked Bloom filter, scalable Bloom filter, Cuckoo filter, Binary Fuse filter, xor filter, HyperLogLog, Theta sketch, Count-Min sketch, Top-K, KLL sketch, DDSketch quantile sketch, MinHash, RIBLT. Zero dependencies.

Implements and evaluates TrieHH and SFP heavy-hitter algorithms under local differential privacy, including full data preprocessing, simulation, and F1-score analysis on Wikipedia Clickstream. Designed for reproducible experiments and privacy–utility benchmarking.

Comparing the Double Space Saving DS to the Heavy Keeper DS for the real world input streams

PoC (Proof of Concept) caching optimization algorithm for graphical tree environment, written entirely in pure C++.

HyperLogLog and Count-Min Sketch in plain Java 21 that count unique and top items in kilobytes, with the observed error tested against the theoretical bounds.