Misra-Gries Frequent-Items Summary: Heavy Hitters with k Counters and a Deterministic Error Bound
Find the heavy hitters of a data stream with only k counters. Each arrival increments a stored counter, fills a free slot or decrements every counter at once. Change k, the skew and the arrival order, and watch the summary stay within N/(k+1) of the true counts. A Blender-built plotting easel and jars, with numerical checks of the error guarantee, the mass identity and the no-miss theorem built in.
The Semantic Web represents everything as subject-predicate-object triples, and this simulator renders a live one as a real 3D force-directed graph: classes, individuals and values repel each other and are pulled together along predicate edges until the layout settles, exactly like a real knowledge-graph visualizer. Flip on RDFS reasoning to watch the reasoner walk the subclass hierarchy and add inferred type edges no one asserted by hand, add random triples to grow the graph, and build a basic-graph-pattern query from the Subject / Predicate / Object dropdowns — using ? as a SPARQL-style wildcard — to see exactly which triples in the store match, the same mechanism a real SPARQL endpoint uses over Linked Data such as DBpedia or Wikidata.
Find the heavy hitters of a data stream with only k counters. Each arrival increments a stored counter, fills a free slot or decrements every counter at once. Change k, the skew and the arrival order, and watch the summary stay within N/(k+1) of the true counts. A Blender-built plotting easel and jars, with numerical checks of the error guarantee, the mass identity and the no-miss theorem built in.
3D · Three.js / WebGL renderer · 60 FPS target · runs fully client-side, no install