BFS, DFS, Dijkstra’s, A*，。
Graphs
Graphs are a powerful and fundamental data abstraction in computer science. They are defined to be a set of vertices and edges and can be used to represent many things, such as network connections, dependencies, image compositions, state machines, and artificial neural networks. It is critical for you understand graphs if you want to pursue a future in computer science - whether that’s in research or industry. We will explore some of the most fundamental graph algorithms today, namely breadth first search (BFS), depth first search (DFS), Dijkstra’s algorithm, and A star algorithm (A*).
Introduction to our Maze Solver
In this lab, we’ll explore how a few graph algorithms behave in the context of mazes, like the one shown below. Note that we will only be using this maze visualization with a select few of the graph algorithms (BFS, DFS, and A*, not Dijkstra’s; we will work with Dijkstra’s algorithm in a different way for this lab).
One way to represent a maze is as an undirected graph. The vertices of such a graph are shown below, with one dimensional (vertex number) coordinates on the left version and (X, Y) coordinates on the right version. If there is no wall between two adjacent vertices, then the corresponding graph has an undirected edge between the vertices. For example, adj(11) would be an iterable containing the integers 12 and 16.
Following the representation method dictated above, both of these mazes could be represented as the graph shown below, where each red circle represents a vertex (also called nodes) and each black line represents an undirected edge between two vertices.
In this lab, you’ll implement a few key graph algorithms using the provided Maze class, which has the following methods of interest:
1
2
3
4
5
6
|
public int N(): Size of the maze [mazes are N x N]
public int V(): Total number of vertices in the maze
public Iterable adj(int v): Returns the neighbors of v
public int toX(int v): Returns the x coordinate of vertex v
public int toY(int v): Returns the y coordinate of vertex v
public int xyTo1D(int x, int y): Returns the vertex number of x, y
|
We’ve provided MazeDepthFirstPaths.java, a version of depth first paths adapted for use with mazes. Later in the lab, we will ask you to implement breadth first paths and a cycle detection algorithm. For those of you who want a deeper understanding of graph algorithms, we’ve also provided an optional challenge to implement the A* shortest paths algorithm (which may prove to be a useful exercise as practice for project 3). For this particular section of the lab, you will be required to modify the following files:
- MazeBreadthFirstPaths.java: Uses BFS to find all paths from a given source, terminating as soon as the target vertex is observed.
- MazeCycles.java: Searches for cycles in the maze. If a cycle is detected, the algorithm terminates.
The following file is optional:
- MazeAStarPath.java: Searches for the shortest path from source to a target using A*, terminating when the target is observed.
These maze solvers should be subclasses of the abstract MazeExplorer class, which has the following fields and methods:
1
2
3
4
5
6
|
public boolean[] marked: Locations to mark in blue
public int[] distTo: Distances to draw over the Maze
public int[] edgeTo: Edges to draw on Maze
public Maze maze: The maze being solved
public void announce(): Call whenever you want the drawing updated
public solve(): Solves the given Maze problem
|
The Maze class has special functionality built in so that it can see your MazeExplorer’s public variables. Specifically, whenever you call announce, it will draw the contents of your MazeExplorer’s marked, distTo, and edgeTo arrays. Make sure to call the announce method any time you want the drawing to be updated.
As an example, try compiling and running TrivialMazeExplorerDemo.java in IntelliJ. Open up the TrivialMazeExplorer.java and TrivialMazeExplorerDemo.java files to understand what’s going on.
If you would like to compile these files without IntelliJ, simply run make or javac *.java in the lab directory with all the Java files. To run a particular demo, say TrivialMazeExplorerDemo, execute java TrivialMazeExplorerDemo in command line.
As a more complex example, try compiling and running DepthFirstDemo. This code was adapted from the DepthFirstPaths class.
Before moving to the next section, try tweaking the parameters of the maze, by editing the maze.config file. There are three different types of mazes (SINGLE_GAP, POPEN_SOLVABLE, and BLANK) to choose from. A % sign at the beginning of a line in the config file comments it out. Feel free to play around with all different types by changing which MazeTypes are commented out.
Breadth First Search
BFS and DFS are very similar. BFS uses a queue (FIFO) to store the fringe, whereas DFS uses a stack (FILO). Naturally, programmers often use recursion for DFS, since we can take advantage of and use the implicit recursive call stack as our fringe. For BFS? There are no implicit structures that we can use. We must use a loop to simulate the expansion of the fringe.
You will need to use a FIFO queue for this part. Luckily, Java’s powerful library already has a Queue interface (a sub-interface of the almighty Collection interface) built in. Read the interface documentation carefully. Hint: see if you can find a familiar Collection-implementing class that also implements this Queue interface. Feel free to use any of them.
You’ll now write a class MazeBreadthFirstPaths.java that extends MazeExplorer. It is highly recommended you look at the code in MazeDepthFirstPaths and use it as inspiration. When you compile and run BreadthFirstDemo.java, you should see your algorithm crawl the graph, locating the shortest path from position (1, 1) to (N, N), stopping as soon as the (N, N) position is found.
You can also use BreadthFirstPaths as inspiration, as well as this video created by Professor Hug showcasing the expected behavior of each class (though there’s a small bug in MazeBreadthFirstPaths that he pointed out during the video).
Depth First Search amp; Cycle Check
In the world of graph theory, there exist many cycle detection algorithms. For example, the union-find algorithm can detect cycle in O(E * logV) time. For today’s exercise, we will use DFS to detect cycles in this maze (an undirected graph) in O(V + E). For every visited vertex v, if there is an adjacent u such that u is already visited and u is not parent of v, then there is a cycle in graph.
For this part of the lab, you’ll write a cycle detection algorithm. When you compile and run CyclesDemo, you should see your algorithm crawl the graph. If it identifies any cycles, it should connect the vertices of the cycle using purple lines (by setting values in the edgeTo array and calling announce()) and terminate immediately. The vertices of the part of the graph that has been traversed to find the cycle should also be drawn, but there should be no edges connecting the part of the graph that doesn’t contain a cycle. The only edges that should be drawn are the ones connecting the cycle. Consider using another data structure to keep track of edges until a cycle is detected and edgeTo is updated, in order to avoid displaying edges that are not part of the cycle. As a reminder, you can change the maze type by making edits to your maze.config file, allowing you to test on a maze with cycles.
Recall from last section, you can use either recursion or a Stack class for DFS. If you decide to go with latter, you need to look up the Java Stack class, or use some linear structure in a FILO fashion.
Dijkstra’s Algorithm
Dijkstra’s algorithm is useful when we want to find the shortest paths from a starting vertex to all vertices in the graph, effectively finding the ‘shortest paths tree’ (SPT).
You will be implementing Dijkstra’s algorithm in the file Graph.java. Take a look at that class and familiarize yourself with how the graph is being represented. For this method, assume that each edge in the graph has a weight field that is a positive integer, which will be represented by the Edge object’s edgeWeight field.
For this exercise, we will not ask you to reconstruct the paths in your algorithm. However, implementing this functionality is good practice and may be something you want to try out on your own!
Hint 1: At a certain point in Dijkstra’s algorithm, you have to change the value of nodes in the fringe. Java’s priority queue does not support this operation directly, but you can add a new entry into the priority queue that contains the updated value (and will always be dequeued before any previous entries). Then, by tracking which nodes have been visited already, you can simply ignore any copies after the first copy dequeued.
Hint 2: Adding the vertices to our priority queue fringe directly won’t be enough. Our vertices are Integers, so our priority queue will order them by their natural ordering. Is this what we want? If not, is there a way we can change how to order the items in the priority queue?
Runtime
When implemented properly using a binary heap, Dijkstra’s algorithm should run in O((|V| + |E|) * log |V|) time. This is because at all times our heap size remains a polynomial factor of |V| (even with lazy removal, our heap size never exceeds |V|^2), and we make at most |V| dequeues and |E| updates requiring heap operations.
For connected graphs, the runtime can be simplified to O(|E| * log |V|), as the number of edges must be at least |V|-1. Using alternative implementations of the priority queue can lead to increased or decreased runtimes.
A* (optional, but worth a read)
In graph theory, determining the shortest path between two nodes is one of the most common and important questions asked. This problem has many real world examples, with shortest route between two cities as one of the most overused in computer science. Dijkstra’s algorithm is the most basic shortest path algorithm and can find the shortest path between two points assuming no negative edge weights. Dijkstra’s is very similar to BFS, for we can replace the queue in BFS with a priority queue (with some more minor tweaks) and end up with Dijkstra’s!
However, Dijkstra’s algorithm is a uniform-cost search. If we want to find the shortest path from SF to NYC, Dijkstra’s will expand in all directions centered at SF, meaning the traversal will reach Seattle before even getting close to the East coast. It can explore in the wrong direction and end up wasting time doing unnecessary work. However, we can make it “smarter” by giving it a heuristic.
Introducing A* (“A star”) search! A* is the state-of-the-art shortest path finding algorithm (given that the programmer provides a good heuristic, such as the Manhattan distance to the destination). Let’s go back to the SF to NYC path finding analogy: with this new heuristic, A* will prioritize moving to the East first, since our heuristic will tell it that moving straight up toward Seattle (going to Canada?) is bad compared to moving toward Denver (closer to NYC).
Here is a nice visualization for BFS, DFS, Dijkstra’s algorithm, and A* algorithm. We highly recommend playing around with it to improve your understanding of these most basic graph algorithms.
For this part of the lab, you’ll implement the A* algorithm. When you compile and run AStarDemo, you should see your algorithm crawl the graph, seeking the shortest path from (1, 1) to (N, N). Unlike BFS, the algorithm should take into account the target vertex.
For your heuristic h(v), we recommend that you use the Manhattan distance, which can be simply expressed as:
1
|
Math.abs(sourceX - targetX) + Math.abs(sourceY - targetY);
|
Experiment with different graph types and heuristics to see how they behave.
Submission
- You need to submit MazeBreadthFirstPaths.java, MazeCycles.java, and Graph.java with Dijkstra’s algorithm implemented.
- MazeAStarPath.java is optional.
- Make sure to run BreadthFirstDemo and CyclesDemo and that your code functions correctly before you submit