// LLP-BellmanFord: when some G[j] is improvable, run one Jacobi-style relaxation.

class LLPBellmanFord {
  int[] LLPBellmanFord(set<int>[] pre, int[][] w) {
    int[] G = infinity;
    G[0] = 0;
    forbidden (j) :
      exists i in pre[j] : G[j] > G[i] + w[i][j]
    =>
      advance :
        forall k in [0..n] :
          G[k] = min(G[k], min i in pre[k] : G[i] + w[i][k]);
    return G;
  }
}