help in the Donalds B. Johnson's algorithm, i cannot understand the pseudo code (PART II)

Question

i cannot understand a certain part of the paper published by Donald Johnson about finding cycles (Circuits) in a graph.

More specific i cannot understand what is the matrix Ak which is mentioned in the following line of the pseudo code :

Ak:=adjacency structure of strong component K with least vertex in subgraph of G induced by {s,s+1,....n};

to make things worse some lines after is mentins " for i in Vk do " without declaring what the Vk is...

As far i have understand we have the following: 1) in general, a strong component is a sub-graph of a graph, in which for every node of this sub-graph there is a path to any node of the sub-graph (in other words you can access any node of the sub-graph from any other node of the sub-graph)

2) a sub-graph induced by a list of nodes is a graph containing all these nodes plus all the edges connecting these nodes. in paper the mathematical definition is " F is a subgraph of G induced by W if W is subset of V and F = (W,{u,y)|u,y in W and (u,y) in E)}) where u,y are edges , E is the set of all the edges in the graph, W is a set of nodes.

3)in the code implementation the nodes are named by integer numbers 1 ... n.

4) I suspect that the Vk is the set of nodes of the strong component K.

now to the question. Lets say we have a graph G= (V,E) with V = {1,2,3,4,5,6,7,8,9} which it can be divided into 3 strong components the SC1 = {1,4,7,8} SC2= {2,3,9} SC3 = {5,6} (and their edges)

Can anybody give me an example for s =1, s= 2, s= 5 what if going to be the Vk and Ak according to the code?

The pseudo code is in my previous question in Understanding the pseudocode in the Donald B. Johnson's algorithm

and the paper can be found at Understanding the pseudocode in the Donald B. Johnson's algorithm

thank you in advance

Answer 1

It works! In an earlier iteration of the Johnson algorithm, I had supposed that A was an adjacency matrix. Instead, it appears to represent an adjacency list. In that example, implemented below, the vertices {a, b, c} are numbered {0, 1, 2}, yielding the following circuits.

Addendum: As noted in this proposed edit and helpful answer, the algorithm specifies that unblock() should remove the element having the value w, not the element having the index w.

list.remove(Integer.valueOf(w));

Sample output:

0 1 0
0 1 2 0
0 2 0
0 2 1 0
1 0 1
1 0 2 1
1 2 0 1
1 2 1
2 0 1 2
2 0 2
2 1 0 2
2 1 2

By default, the program starts with s = 0; implementing s := least vertex in V as an optimization remains. A variation that produces only unique cycles is shown here.

import java.util.ArrayList;
import java.util.Arrays;
import java.util.List;
import java.util.Stack;

/**
 * @see http://dutta.csc.ncsu.edu/csc791_spring07/wrap/circuits_johnson.pdf
 * @see https://stackoverflow.com/questions/2908575
 * @see https://stackoverflow.com/questions/2939877
 * @see http://en.wikipedia.org/wiki/Adjacency_matrix
 * @see http://en.wikipedia.org/wiki/Adjacency_list
 */
public final class CircuitFinding {

    final Stack<Integer> stack = new Stack<Integer>();
    final List<List<Integer>> a;
    final List<List<Integer>> b;
    final boolean[] blocked;
    final int n;
    int s;

    public static void main(String[] args) {
        List<List<Integer>> a = new ArrayList<List<Integer>>();
        a.add(new ArrayList<Integer>(Arrays.asList(1, 2)));
        a.add(new ArrayList<Integer>(Arrays.asList(0, 2)));
        a.add(new ArrayList<Integer>(Arrays.asList(0, 1)));
        CircuitFinding cf = new CircuitFinding(a);
        cf.find();
    }

    /**
     * @param a adjacency structure of strong component K with
     * least vertex in subgraph of G induced by {s, s + 1, n};
     */
    public CircuitFinding(List<List<Integer>> a) {
        this.a = a;
        n = a.size();
        blocked = new boolean[n];
        b = new ArrayList<List<Integer>>();
        for (int i = 0; i < n; i++) {
            b.add(new ArrayList<Integer>());
        }
    }

    private void unblock(int u) {
        blocked[u] = false;
        List<Integer> list = b.get(u);
        for (int w : list) {
            //delete w from B(u);
            list.remove(Integer.valueOf(w));
            if (blocked[w]) {
                unblock(w);
            }
        }
    }

    private boolean circuit(int v) {
        boolean f = false;
        stack.push(v);
        blocked[v] = true;
        L1:
        for (int w : a.get(v)) {
            if (w == s) {
                //output circuit composed of stack followed by s;
                for (int i : stack) {
                    System.out.print(i + " ");
                }
                System.out.println(s);
                f = true;
            } else if (!blocked[w]) {
                if (circuit(w)) {
                    f = true;
                }
            }
        }
        L2:
        if (f) {
            unblock(v);
        } else {
            for (int w : a.get(v)) {
                //if (v∉B(w)) put v on B(w);
                if (!b.get(w).contains(v)) {
                    b.get(w).add(v);
                }
            }
        }
        v = stack.pop();
        return f;
    }

    public void find() {
        while (s < n) {
            if (a != null) {
                //s := least vertex in V;
                L3:
                circuit(s);
                s++;
            } else {
                s = n;
            }
        }
    }
}

Answer 2

The following variation produces unique cycles. Based on this example, it is adapted from an answer supplied by @user1406062.

Code:

import java.util.ArrayList;
import java.util.Arrays;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.Stack;

/**
 * @see https://en.wikipedia.org/wiki/Johnson%27s_algorithm
 * @see https://stackoverflow.com/questions/2908575
 * @see https://stackoverflow.com/questions/2939877
 * @see http://en.wikipedia.org/wiki/Adjacency_matrix
 * @see http://en.wikipedia.org/wiki/Adjacency_list
 */
public final class CircuitFinding {

    final Stack<Integer> stack = new Stack<Integer>();
    final Map<Integer, List<Integer>> a;
    final List<List<Integer>> b;
    final boolean[] blocked;
    final int n;
    Integer s;

    public static void main(String[] args) {
        List<List<Integer>> a = new ArrayList<List<Integer>>();
        a.add(new ArrayList<Integer>(Arrays.asList(1, 2)));
        a.add(new ArrayList<Integer>(Arrays.asList(0, 2)));
        a.add(new ArrayList<Integer>(Arrays.asList(0, 1)));
        CircuitFinding cf = new CircuitFinding(a);
        cf.find();
    }

    /**
     * @param a adjacency structure of strong component K with least vertex in
     * subgraph of G induced by {s, s + 1, n};
     */
    public CircuitFinding(List<List<Integer>> A) {
        this.a = new HashMap<Integer, List<Integer>>(A.size());
        for (int i = 0; i < A.size(); i++) {
            this.a.put(i, new ArrayList<Integer>());
            for (int j : A.get(i)) {
                this.a.get(i).add(j);
            }
        }
        n = a.size();
        blocked = new boolean[n];
        b = new ArrayList<List<Integer>>();
        for (int i = 0; i < n; i++) {
            b.add(new ArrayList<Integer>());
        }
    }

    private void unblock(int u) {
        blocked[u] = false;
        List<Integer> list = b.get(u);
        for (int w : list) {
            //delete w from B(u);
            list.remove(Integer.valueOf(w));
            if (blocked[w]) {
                unblock(w);
            }
        }
    }

    private boolean circuit(int v) {
        boolean f = false;
        stack.push(v);
        blocked[v] = true;
        L1:
        for (int w : a.get(v)) {
            if (w == s) {
                //output circuit composed of stack followed by s;
                for (int i : stack) {
                    System.out.print(i + " ");
                }
                System.out.println(s);
                f = true;
            } else if (!blocked[w]) {
                if (circuit(w)) {
                    f = true;
                }
            }
        }
        L2:
        if (f) {
            unblock(v);
        } else {
            for (int w : a.get(v)) {
                //if (v∉B(w)) put v on B(w);
                if (!b.get(w).contains(v)) {
                    b.get(w).add(v);
                }
            }
        }
        v = stack.pop();
        return f;
    }

    public void find() {
        s = 0;
        while (s < n) {
            if (!a.isEmpty()) {
                //s := least vertex in V;
                L3:
                for (int i : a.keySet()) {
                    b.get(i).clear();
                    blocked[i] = false;
                }
                circuit(s);
                a.remove(s);
                for (Integer j : a.keySet()) {
                    if (a.get(j).contains(s)) {
                        a.get(j).remove(s);
                    }
                }
                s++;
            } else {
                s = n;
            }
        }
    }
}

Output:

0 1 0
0 1 2 0
0 2 0
0 2 1 0
1 2 1

All cycles, for reference:

0 1 0
0 1 2 0
0 2 0
0 2 1 0
1 0 1
1 0 2 1
1 2 0 1
1 2 1
2 0 1 2
2 0 2
2 1 0 2
2 1 2

Answer 3

I had sumbitted an edit request to @trashgod's code to fix the exception thrown in unblock(). Essentially, the algorithm states that the element w (which is not an index) is to be removed from the list. The code above used list.remove(w), which treats w as an index.

My edit request was rejected! Not sure why, because I have tested the above with my modification on a network of 20,000 nodes and 70,000 edges and it doesn't crash.

I have also modified Johnson's algorithm to be more adapted to undirected graphs. If anybody wants these modifications please contact me.

Below is my code for unblock().

private void unblock(int u) {
    blocked[u] = false;
    List<Integer> list = b.get(u);
    int w;
    for (int iw=0; iw < list.size(); iw++) {
        w = Integer.valueOf(list.get(iw));
        //delete w from B(u);
        list.remove(iw);
        if (blocked[w]) {
            unblock(w);
        }
    }
}

Answer 4

@trashgod, your sample output contains cycle which are cyclic permutation. For example 0-1-0 and 1-0-1 are same Actually the output should contains only 5 cycle i.e. 0 1 0, 0 2 0, 0 1 2 0, 0 2 1 0, 1 2 1,

Johnson paper explain what a cycle is: 'Two elementary circuits are distinct if one is not a cyclic permutation of the other. ' One can also check wolfram page: This also output 5 cycle for the same input.

http://demonstrations.wolfram.com/EnumeratingCyclesOfADirectedGraph/