Edge Token
A representation scheme in graph neural networks that treats edges as the basic processing unit, mapping each edge directly to an independent feature vector that can incorporate endpoint coordinates, demands, distances and constraints. It aligns naturally with combinatorial optimization problems whose decision space consists of edges.
Timeline (last 90 days)
RELAGRAM encodes CVRP instances as edge tokens (Edge Tokens) and uses a small Transformer network to repeatedly update the latent representations of the edge tokens.
Edge token encoding maps each edge (i, j) directly to an independent feature vector, which can incorporate attributes such as endpoint coordinates, demands, distances, and capacity constraints.
RELAGRAM's edge token encoding is not tied to the total number of nodes, and the parameter count of the recursive Transformer is independent of instance scale, enabling it to achieve cross-scale transfer.
All Facts (3)
RELAGRAM encodes CVRP instances as edge tokens (Edge Tokens) and uses a small Transformer network to repeatedly update the latent representations of the edge tokens.
50%UnverifiedEdge token encoding maps each edge (i, j) directly to an independent feature vector, which can incorporate attributes such as endpoint coordinates, demands, distances, and capacity constraints.
50%UnverifiedRELAGRAM's edge token encoding is not tied to the total number of nodes, and the parameter count of the recursive Transformer is independent of instance scale, enabling it to achieve cross-scale transfer.
50%