12.1 MST practice
a)
After the first step: After the second step:
b)
c)
12.3 Uniqueness of MSTs
a)
We can get MST from Kruskal’s algorithm, and from Prim’s algorithmnot unique MST
Link to original
b)
Since is a spanning tree, adding an extra edge creates exactly one cycle: Assume connects and . There exists a path from to through the spanning tree (without using ). If we add to this path, we get a cycle.
We claim this cycle to be . There must be an edge on , because otherwise would be fully contained in , which contradicts the fact that is a spanning tree. Moreover, could not be , because . So and are distinct. Since all weights are distinct and is the minimum weight, we have
c)
We replace with on . We first add to and get a cycle , then we remove another edge on this cycle, such that the remaining structure is still a spanning tree (every vertices are still connected and no more cycles). Now we’ve reduced the total weight of by (not ). This leads to a contradiction to our assumption that is a minimum spanning tree. Therefore, the minimum spanning tree of with weight function is unique.
d)
There is one unique minimum spanning tree , however, the weights are not distinct.Not unique weight of MST
Link to original
12.4 Counting Minimum Spanning Trees With Identical Edge Weights
a)
Contracting an edge just “glues together” its two endpoints. Since is a tree, this operation does not create cycles or disconnect the graph. So the contracted tree is still a spanning tree of the contracted graph . Now we show that is minimal. Suppose we uncontract (separate , and add back the original edges). The total weight increases by exactly .If there were a spanning tree of cheaper than , then uncontracting it would give a spanning tree cheaper than , which contradicts the fact that is minimal.
b)
We partition into weight classes. we define An edge is redundant if it lies in a class of size at least . ( with ) Let . In , the only weight class whose size changes is , which becomes size . Case 1: , then the redundancy drops by 2. because now has size of 1 and does not contribute to the overall redundancy. Case 2: , then the redundancy drops by 1. Therefore, the number of redundant edges in is at most Hence, is -redundant for some
c)
We prove by induction Claim: If is connected and k-redundant, it has at most distinct MSTs. Base: : If , then all edges are pairwise distinct, so MST is distinct (result from 12.3). The number of distinct MST is
Step: we assume if is connected and k-redundant, it has at most distinct MSTs Consider a (k+1)-redundant : we pick a redundant edge We denote the number of distinct MSTs that contains with We denote the number of distinct MSTs that does not contains with
case 1 by (a) the contracted tree is an MST of , so by (b), is at most ((k+1)-1)-redundant. So we have case 2 Consider . Any MST with is still a spanning tree of . Hence we have: we know deleting a redundant edge decreases the number of redundant edges by at least 1 so Conclusion the number of distinct MSTs We’ve now shown that there are at most distinct MSTs.
The claim is proved through induction.