The Beck–Fiala bound #
Adapted for Lean Pool by changing module paths and selecting explicit imports.
If every vertex of a finite hypergraph belongs to at most t edges, each column of its
incidence matrix has Euclidean norm at most √t. The scaled Komlós bound therefore gives
an edge discrepancy bound of 36 * √t.
Hypergraph.incidenceMatrix has rows indexed by edges and columns indexed by vertices.
Hypergraph.discrepancy_incidenceMatrix_le proves the matrix bound;
Hypergraph.exists_isColouring_forall_abs_finsum_le states it as a vertex colouring.
A hypergraph with finitely many vertices has finitely many edges.
The incidence matrix of a hypergraph with finitely many vertices: its rows are indexed by
the edges, its columns by the vertices, and an entry is 1 when the vertex lies in the edge
and 0 otherwise.
Equations
Instances For
The squared entries of a column of the incidence matrix count the edges containing the vertex.
The Beck–Fiala conjecture, with constant 36: if every vertex of a hypergraph lies in at
most t edges, then its incidence matrix has discrepancy at most 36 * √t.
The Beck–Fiala conjecture, colouring form: if every vertex of a hypergraph lies in at
most t edges, then some colouring of the vertices gives every edge discrepancy at most
36 * √t.