A graph homomorphism from to is a function that sends every edge of to an edge of .
For finite graphs , the homomorphism density is the probability that a uniformly random map is a graph homomorphism. For bipartite graphs with prescribed vertex classes, the random map is chosen separately and uniformly into the corresponding classes.
The Sidorenko conjecture states that every bipartite graph satisfies
for every bipartite host graph . Thus a random map is at least as likely to preserve every edge as the heuristic that treats the edge constraints independently predicts.
Every tree satisfies the Sidorenko conjecture. If a bipartite host graph has edge density , then a uniformly random bipartition-respecting map from a -vertex tree into the host is a graph homomorphism with probability at least .

Articles by others on the same topic (1)