peredge

Peredge is a term used in computing to describe a per-edge data processing paradigm in graph-based systems. In this view, computation focuses on individual edges rather than nodes, enabling localized transformations and parallel execution along the graph's edges. The term is used primarily in discussions of edge-centric graph processing and streaming graphs where edge updates carry the primary meaning.

Formalization: For a graph G=(V,E) with edge features, a peredge operation is a function F that assigns to each edge e in E an output from a space X, possibly depending on the edge's own features and adjacent node states. This enables per-edge message passing, feature updating, or routing decisions that propagate along paths. In practice, per-edge models are often implemented by applying the operation independently to each edge and aggregating results at incident nodes or across the graph through selective communication.

Applications and examples: Per-edge processing is used in streaming graph analytics, where edge updates arrive in real time, and in certain graph neural networks that emphasize edge features. It also appears in edge-computing scenarios and in routing or flow optimization problems where decisions are tied to individual connections rather than entire nodes.

Status and terminology: Peredge is not a standardized term with a single canonical definition. It is sometimes used informally to contrast edge-centric with node-centric approaches. Related concepts include edge-centric graph processing, per-edge features, and graph neural networks. See also edge computing and edge-centric graph processing for neighboring ideas.