[44] | 1 | package cecj.app.othello; |
---|
| 2 | |
---|
| 3 | import ec.EvolutionState; |
---|
| 4 | import ec.Individual; |
---|
| 5 | import ec.util.Parameter; |
---|
| 6 | import ec.vector.DoubleVectorIndividual; |
---|
| 7 | import games.BoardGame; |
---|
| 8 | import games.Player; |
---|
| 9 | import cecj.eval.TDLImprover; |
---|
| 10 | import games.scenarios.SelfPlayTDLScenario; |
---|
| 11 | |
---|
| 12 | public class OthelloTDLImprover implements TDLImprover { |
---|
| 13 | |
---|
| 14 | private static final int WPC_LENGTH = 64; |
---|
| 15 | |
---|
| 16 | private static final String P_REPEATS = "repeats"; |
---|
| 17 | private static final String P_RANDOMNESS = "randomness"; |
---|
| 18 | private static final String P_LEARNING_RATE = "learning-rate"; |
---|
| 19 | |
---|
| 20 | private int repeats; |
---|
| 21 | private double randomness; |
---|
| 22 | private double learningRate; |
---|
| 23 | |
---|
| 24 | public void setup(EvolutionState state, Parameter base) { |
---|
| 25 | Parameter randomnessParam = base.push(P_RANDOMNESS); |
---|
| 26 | randomness = state.parameters.getDoubleWithDefault(randomnessParam, null, 0.1); |
---|
| 27 | |
---|
| 28 | Parameter learningRateParam = base.push(P_LEARNING_RATE); |
---|
| 29 | learningRate = state.parameters.getDoubleWithDefault(learningRateParam, null, 0.01); |
---|
| 30 | |
---|
| 31 | Parameter repeatsParam = base.push(P_REPEATS); |
---|
| 32 | repeats = state.parameters.getIntWithDefault(repeatsParam, null, 10); |
---|
| 33 | } |
---|
| 34 | |
---|
| 35 | public void improve(EvolutionState state, Individual ind) { |
---|
| 36 | Player player = new OthelloPlayer(getWPC(state, ind)); |
---|
| 37 | BoardGame game = new OthelloGame(new OthelloBoard()); |
---|
| 38 | SelfPlayTDLScenario selfPlayScenario = new SelfPlayTDLScenario(state.random[0], player, |
---|
| 39 | randomness, learningRate); |
---|
| 40 | |
---|
| 41 | for (int r = 0; r < repeats; r++) { |
---|
| 42 | game.reset(); |
---|
| 43 | selfPlayScenario.play(game); |
---|
| 44 | } |
---|
| 45 | } |
---|
| 46 | |
---|
| 47 | public void prepareForImproving(EvolutionState state, Individual ind) { |
---|
| 48 | double[] wpc = getWPC(state, ind); |
---|
| 49 | for (int i = 0; i < WPC_LENGTH; i++) { |
---|
| 50 | wpc[i] = 0.0; |
---|
| 51 | } |
---|
| 52 | } |
---|
| 53 | |
---|
| 54 | private double[] getWPC(EvolutionState state, Individual ind) { |
---|
| 55 | if (!(ind instanceof DoubleVectorIndividual)) { |
---|
| 56 | state.output.error("Othello players should be represented by floats vectors\n"); |
---|
| 57 | } |
---|
| 58 | |
---|
| 59 | double[] wpc = ((DoubleVectorIndividual) ind).genome; |
---|
| 60 | if (wpc.length != WPC_LENGTH) { |
---|
| 61 | state.output.error("Players WPC vectors length should be 64\n"); |
---|
| 62 | } |
---|
| 63 | |
---|
| 64 | return wpc; |
---|
| 65 | } |
---|
| 66 | } |
---|