Algorithm Exam Help

Introduction To Algorithms
Graph Problems
• Graph reachability by BFS or DFS in O(|E|) time
• Graph exploration/connected components via Full-BFS or Full-DFS
• Topological sort / Cycle detection via DFS
• Negative-weight cycle detection via Bellman-Ford
• Single Source Shortest Paths (SSSP)
Restrictions SSSP
Algorithm
Graph Weights Name Running Time
O(·)
DAG Any DAG |V | + |E|
Relaxation |V | + |E|
General Unweighted BFS
General Non- Dijkstra |V | log |V | + |
negative Bellman-Ford E|
General Any |V | · |E|
• All Pairs Shortest Paths (APSP)

– Run a SSSP algorithm |V | times
– Johnson’s solves APSP with negative weights in O(|V |2 log |V | + |V
||E|)
– Graph Problem Strategies
• Be sure to explicitly describe a graph in terms of problem parameters
• Convert problem into finding a shortest path, cycle, topo. sort, conn.
comps., etc.
© 2022 All Rights Reserved . Programming Exam Help

• May help to duplicate graph vertices to encode additional
information
• May help to add auxiliary vertices/edges to graph
• May help to pre-process the graph (e.g., to remove part of the
graph)
Graph Problem Common Mistakes
• Define your graphs! Specify vertices, edges, and weights clearly (and
count them!)
– (e.g., construct graph G = (V, E) with a
vertex for each... and a directed edge (u, v)
with weight w for each...)
• State the problem you are solving, not just the
algorithm you use to solve it
– (e.g., solve SSSP from s by running DAG
Relaxation...)
• Connect the graph problem you solve back to the
original problem
– (e.g., the weight of a path from s to t in G corresponds to the
sum of tolls paid along a driving route, so a path of minimum
weight corresponds to a route minimizing tolls)
Problem 1. Counting Blobs

An image is a 2D grid of black and white square pixels where each
white pixel is contained in a
blob. Two white pixels are in the same blob if they share an edge of the
grid. Black pixels are not contained in blobs. Given an n × m array
representing an image, describe an O(nm)-time algorithm to count the
number of blobs in the image.
Solution: Construct a graph G with a vertex per white pixel, with an
undirected edge between two vertices if the pixels associated with them are
both white and share an edge of the grid. This graph has size at most
O(nm) vertices and at most O(nm) edges (as pixels share edges with at
most four other pixels), so can be constructed in O(nm) time. Each
connected component of this graph corresponds to a blob, so run Full-BFS
or Full-DFS to count the number of connected components in G in O(nm)
time
Problem 2. Unicycles
Given a connected undirected graph G = (V, E) with strictly positive

weights w : E → Z+ where |E| = |V |, describe an O(|V |)-time algorithm to
determine a path from vertex s to vertex t of minimum weight.
Solution: Given two vertices in a weighted tree containing only

positive weight edges, there is a unique simple path between them
which is also the minimum weight path. A depth-first search from any
source vertex s in the tree results in a directed DFS tree in O(|V |)
time (since
|E| = |V | − 1). Then relaxing edges in topological sort order of the
directed DFS tree computes minimum weight paths from s in O(|V |)
time. Since G has one cycle, our strategy will be to break the cycle by
removing an edge, and then compute the minimum weight path from s
to t in the resultant tree.
First, we find the vertex v closest to s on the cycle by running depth-
first search from s in O(|V |) time (since |E| = |V |). One edge e1 of
the cycle will not be in the tree returned by DFS (a back edge to v),
with the other edge of the cycle incident to v being a single outgoing
DFS tree edge e2 . If s is on the cycle, v = s; otherwise the unique
path from s to v does not
contain e1 or e2 . A shortest path from s to t cannot traverse both edges e1
and e2 , or else the path would visit v at least twice, traversing a cycle of
positive weight. Removing either e1 or e2 results in a tree, at least one of
which contains the minimum weight path from s to t. Thus, find the
minimum weight path from s to t in each tree using the algorithm
described above, returning the minimum of the two in O(|V |) time.
Problem 3. Doh!-nut
Momer has just finished work at the FingSprield power plant at
location p, and needs to drive to his home at location h. But along the
way, if his driving route ever comes within driving distance k of a
doughnut shop, he will stop and eat doughnuts, and his wife, Harge,
will be angry. Momer knows the layout of Fing Sprield, which can be
modeled as a set of n locations, with two-way roads of known driving
distance connecting some pairs of locations (you may assume that no
location is incident to more than five roads), as well as the locations
of the d doughnut shops in the city. Describe an O(n log n)-time
algorithm to find the shortest driving route from the power plant back
home that avoids driving within driving distance k of a doughnut shop
(or determine no such route exists).
Solution: Construct a graph G with a vertex for each of the n city
locations, and an undirected edge between two locations if there is a
road connecting them, with each edge weighted by the positive length
of its corresponding road. The degree of each vertex is bounded by a
constant (i.e., 5), so the number of edges in G is O(n). First, we
identify vertices that are within driving distance k of a doughnut shop
location: create an auxiliary vertex x with a 0-weight outgoing edge
from x to every doughnut shop location, and run Dijkstra from x.
Remove every vertex from the graph whose shortest path from x is
less

than or equal to k, resulting in graph G0 ⊂ G. If either p or h are not
in G0, then no route exists. Otherwise, run Dijkstra from p in G0. If no
path exists to h, then no valid route exists. Otherwise, Dijkstra finds a
shortest path from p to h, so return it (via parent pointers). This
algorithm runs Dijkstra twice. Since the size of either graph is O(|V
|), Dijkstra runs in O(|V | log |V |) = O(n log n) time (e.g. using a
binary heap to implement a priority queue).
Problem 4. Long Shortest Paths
Given directed graph G = (V, E) having arbitrary edge weights w :
E → Z and two vertices s, t ∈ V , describe an O(|V |3)-time
algorithm to find the minimum weight of any path from s to t
containing at least |V | edges.
Solution: Our strategy will compute intermediate values for each vertex
v∈V:
1. the minimum weight w1(v) of any path from s to v using exactly |V
| edges, and then
2. the minimum weight w2(v) of any path from v to t using any
number of edges.
First, to compute (1), we make a duplicated graph similar to Bellman-
Ford, but without edges corresponding to remaining at a vertex.
Specifically, construct a graph G1 with
|V | + 1 vertices for each vertex v ∈ V : vertex vk for k ∈ {0, . . . , |
V |} representing reaching
v from s along a path containing k edges; and
• |V | edges for each edge (u, v) ∈ E: edge (uk−1, vk ) of the same
weight for k ∈ {1, . . . , |V |}.
Now a path in G1 from s0 to v|V | for any v ∈ V corresponds to a path

from s to v in G through exactly |V | edges. So solve SSSPs in G1
from s0 to compute the minimum weight of paths to each vertex
|(|V | + |E|)) = O(|V |3), so we can solve SSSP on G1 via DAG
relaxation in O(|V |3) time.
Second, to compute (2), we make a new graph G2 from G where
every edge is reversed. Then every path to t in G corresponds to a
path in G2 from t, so compute SSSPs from t in G2 to find the
minimum weight of any path from v to t in G using any number of
edges, which can be done in O(|V ||E|) = O(|V |3) time using
Bellman-Ford.
Once computed, finding the minimum sum of w1 (v) + w2 (v) over all
vertices v ∈ V will provide the minimum weight of any path from s
to t containing at least |V | edges, since every such path can be
decomposed into its first |V | edges and then the remainder. This loop
takes O(|V |) time, so the algorithm runs in O(|V |3 ) time in total.

Algorithm Exam Help

Uploaded by

Algorithm Exam Help

Uploaded by

Introduction To Algorithms

• All Pairs Shortest Paths (APSP)

© 2022 All Rights Reserved . Programming Exam Help

Problem 1. Counting Blobs

Given a connected undirected graph G = (V, E) with strictly positive

Solution: Given two vertices in a weighted tree containing only

© 2022 All Rights Reserved . Programming Exam Help

Now a path in G1 from s0 to v|V | for any v ∈ V corresponds to a path

© 2022 All Rights Reserved . Programming Exam Help

You might also like