Nested loop in 2D array stores same value across rows

[Nicias]

Hi all,

I’m trying to run a simulation where 2 players out of N meet at each iteration t. I use a 1D array to store some properties of these players and use another 2D array to store these properties for all N players and all T iterations, since players don’t need to remember them. Problem is that I don’t know how but I’m only storing in this 2D array last values of the 1D array, not the whole record. For instance, if player 1 starts with some attribute named “belief” whose value is 0.9, one record after 5 iterations could be: 0.9, 0.45, 0.67, 0.93, 0.87. But in my 2D array I just can read 0.87 five times. However, when running the program, I can see that player 1 is really changing this value of “belief” and how it takes 0.9, 0.45, 0.67 and so on… Just I can’t manage to store this whole record and I only save T times the same last value of it.

Here is the code:

public void simulate()
   {      
      Uniform uniformAgent = new Uniform( 0, N-1, seed );      

      for ( int t = 0; t < T; t++ )
      {
         int a1 = uniformAgent.nextInt();
         System.out.println( "Agent 1 is " + Integer.toString( a1 ) );
         int a2 = uniformAgent.nextInt();
         while ( a1 == a2 )
         {
            a2 = uniformAgent.nextInt();
         }
         System.out.println( "Agent 2 is " + Integer.toString( a2 ) );     
                 
         agents[a1].play();
         agents[a2].play();

         for ( int n = 0; n < N; n++ )
         {
            actions[n][t] = agents[n].getAction();
         }

         agents[a1].setOthersAction( agents[a2].getAction() );
         agents[a2].setOthersAction( agents[a1].getAction() );
         agents[a1].updateReturns();
         agents[a2].updateReturns();

         double pay1 = agents[a1].getPayoff().getPayoff();
         double pay2 = agents[a2].getPayoff().getPayoff();
         System.out.println( "Payoff of agent1 is " +
                 Double.toString( pay1 ) );
         System.out.println( "Payoff of agent2 is " +
                 Double.toString( pay2 ) );

         System.out.println( "Belief of agent1 is "
                 + Double.toString( agents[a1].getBelief().getBelief() ) );
         System.out.println( "Belief of agent2 is "
                 + Double.toString( agents[a2].getBelief().getBelief() ) );

         for ( int n = 0; n < N; n++ )
         {
            beliefs[n][t] = agents[n].getBelief();
            System.out.println( beliefs[n][t].getBelief() );
            payoffs[n][t] = agents[n].getPayoff();
         }
      }
} // end simulate method

Thanks in advance.

[SJF]

Not really following this here:

for ( int n = 0; n < N; n++ )
         {
            beliefs[n][t] = agents[n].getBelief();
            System.out.println( beliefs[n][t].getBelief() );
            payoffs[n][t] = agents[n].getPayoff();
         }

what is beliefs? an agent? a double?
If beliefs is a double... beliefs[n][t].getBeliefs() makes no sense.
if beliefs is an agent... I would think you'd set beliefs[n][t].setBelief(agents[n].getBelief()) or some similar (not to mention the overhead of a 2D object array o.O )

if beliefs is a 2D array of agents... "beliefs[n][t] =" is a pass by reference and could be very ... cumbersome

[Nicias]

Design consists of a Belief class, which inherits directly from Object. Belief class has a double field, belief, which is accessed with getBelief() method.
Agent class has a Belief object as a field and method getBelief() returns a Belief object.
Finally, Breen class has a 1D array of Agent objects. Breen class has simulate() method I show here. Breen also has a 2D Belief array to store each Belief object of each Agent object in each iteration of “t”.

So agent[n].getBelief() returns a reference to a Belief object of nth agent and line 44 here should store this reference in beliefs[n][t].

In general, the code makes this:

for each time “t”;
choose two random agents which are not the same;
update some values of these two agents through play() method;
get those values updated for all “n” agents;
update the rest of values of these two agents;
store the rest of values of all “n” agents;

The code is performing as expecting in everything. If some agent has belief value of 0.9 at iteration “t=0”, it’s possible to see how it changes while the program is running, through lines 36-39 here. agent[n].getBelief() returns correct and valid values. The matter is that I can’t get to store the whole history. I want to track each value of Belief for each agent “n” for each time “t”, but my code only stores the last value for each agent as many times as I define T, in line 5. While on the prompt I can see how Belief for some agent changes from 0.9 to 0.45 to 0.66 or whatever, later this history isn’t recorded, only the last value of it, so I finish with a row like [0.66, 0.66, 0.66] instead of [0.9, 0.45, 0.66].

I hope this makes my code clearer. The fact is that I have made this before in Matlab, with the same design, and there it runs well and store values properly. But I wanted in Java. Running this in Matlab I can go for lunch, a walk… before it finishes. In Java it just takes seconds or a couple of minutes if I make thousands of iterations. I can’t dream with making that in Matlab.

[Nicias]

Ok. I have got that my program stores some part of the history of "beliefs" in line 43. If it takes values 0.9, 0.37, 0.65, this line of code now stores these values, not 3 times 0.65. I have made it adding a call to getBelief() method Belief class, which gives a double back. Line 43 reads now:

beliefs[n][t] = agents[n].getBelief().getBelief();

Of course, I have also changed definition of array beliefs, to be of double type, instead of Belief class type as before. It seems here I was passing reference of objects, instead of just value, so that every time they changed inside any Agent object, they also changed in 2D array beliefs where all of them are tracked.

However, another problem remains still. I have said now it stores part of the history, because it doesn't store all values. I mean, I create a 2D array named "beliefs" with dimensions NxT, but the output is an array of dimensions NxN.

Maybe it helps if I add fields declaration and constructor, where this array is created, so that someone can help me:

public class Breen {
   private int N, T, seed;
   double BELIEF, HSLASH, B, C, p;
   Uniform uniform;
   double opinions[];
   Agent agents[];
   Action actions[][];
   double beliefs[][];
   Payoff payoffs[][];

// some methods here

public Breen( int _N, int _T, double _p, int _seed, double _belief,
           double _hslash, double _B, double _C )
   {
      this.N = _N;
      this.T = _T;
      this.seed = _seed;
      this.BELIEF = _belief;
      this.HSLASH =_hslash;
      this.B = _B;
      this.C = _C;
      this.p = _p;
      opinions = new double[N];
      agents = new Agent[N];
      uniform = new Uniform( 0.0, 1.0, seed);

      try {
      for ( int i = 0; i < N; i++ )
      {
         opinions[i] = bernoulli();         
         agents[i] = new Agent( BELIEF, HSLASH, opinions[i], B, C );
         
      }

      } catch ( Exception e )
      {
         System.err.println( e.getMessage() + "\n" );
         e.printStackTrace();
      }

      actions = new Action[N][T];
      beliefs = new double[N][T];
      payoffs = new Payoff[N][T];

   } // end constructor method

Thanks in advance.

[Nicias]

Solved. Array was size NxT because of, on the interface of the program, I defined that T took value from same JTextField as N. Now it works. Sorry for any inconveniences.

[Nicias]

This problem is already solved, so some some moderator could mark it as such.