Enviroment.txt

//by: Annahita Doulatshahi
//Trying to solve for multi-arm bandit problem with three different algorithms.

//Game setting: Enviroment object
//Contains {main(),printCurrentResults(int i),printOpening()}
//*******printOpening(): called once in main(), basically prints legend to the remaining outputs
//*******printCurrentResults(int i): prints the current frequency of choosing a specific Bandit-arm for each agent 
//*******main()


public class Environment {
	
	static private int[] score;
	static private int[][] opCount;
	static private double[][][] freqList;
	

	private static void printCurrentResults(int i){
		 System.out.println("\n"+ "Round " + i + ":");
		 System.out.println("Gready Score = " + score[0]);
		 System.out.println("LinearPenalty score = " + score[1]);
		 System.out.println("LinearInactive Score = " + score[2]);
		 System.out.println("Random score = " + score[3]);
		 
		 
		 //print optimal move
		 for(int x=0;x<4;x++){
			 System.out.println("\n Bot "+ x + ": ");
			 System.out.println("________________");
			 System.out.println("Rounds");
			 System.out.println("              1     2     3     4     5     6     7     8     9    10 ");
				 for(int n=0;n<10;n++){
					 System.out.print("BanditArm "+ (n+1) + ": ");
					 System.out.print(String.format("%.1f", freqList[x][0][n]) + "  ");
					 System.out.print(String.format("%.1f", freqList[x][1][n])  + "  ");
					 System.out.print(String.format("%.1f", freqList[x][2][n])  + "  ");
					 System.out.print(String.format("%.1f", freqList[x][3][n]) + "  ");
					 System.out.print(String.format("%.1f", freqList[x][4][n])  + "  "); 
					 System.out.print(String.format("%.1f", freqList[x][5][n])  + "  ");
					 System.out.print(String.format("%.1f", freqList[x][6][n])  + "  ");
					 System.out.print(String.format("%.1f", freqList[x][7][n])  + "  ");
					 System.out.print(String.format("%.1f", freqList[x][8][n])  + "  "); 
					 System.out.print(String.format("%.1f", freqList[x][9][n]) ); 
				 }
		 }
	}
		 
    private static void countFreq(int iter, int round){
    	 for(int x=0;x<4;x++){
			 for(int j=0;j<10;j++){
				 freqList[x][round][j]= ((opCount[x][j] / (double)iter)*100);
			 }    
		}
	}
    
	private static void printOpening(){
		System.out.println("Bot List");
		System.out.println("------------------");
		System.out.println("Bot 0 = o-Greedy");
		System.out.println("Bot 1 = LinearRewardInaction");
		System.out.println("Bot 2 = LinearRewardPenalty");
		System.out.println("Bot 3 = RandomAgent");
		
		System.out.println("------------------");
		System.out.println("Bandit Arm Prob of success ranked sequently by num. ex)Arm 1 has the greatest chance for a reward ");

		
	
	}
	 public static void main(String[] args) {
		 printOpening();
		 
		 score = new int[4];
		 opCount = new int[4][10];
		 freqList = new double[4][10][10];//Three bots,10rounds,10arms
		 Bandit b = new Bandit();

		 LearnAgent[] agents = {new Greedy(), new Linear_RewardPenalty(),new Linear_RewardInaction(),new RandomAgent()};
		 int i,k,reward=0;
		
		 for (i = 0; i < 10000; i++){
			if( (i % 1000 == 0) && (i != 0))countFreq(i, (i/1000));
			 
			 //agent execution on round i
			 for(int j = 0; j < 4 ; j++){
				 k  = agents[j].chooseAction();
				 if((k >= 0) && (k < 11)){
					 opCount[j][k]+=1;
					 reward = b.pullArm(k+1);
					 }
				 else 
				 { //error check
				 System.out.println("error in agent choice index");
				 System.out.println("agent " + j + " error: 0=greedy,1=lrp,2=lri,3=random");
				 System.out.println("action choice = " + k);
				 }
				 
				 agents[j].processReward(reward);
				 score[j] += reward;
			 }
			  
		 }
		 printCurrentResults(i);
	 }

}