update plot + validation

markovDecision.py

@@ -60,6 +60,7 @@ def markovDecision(layout : list, circle : bool):
 # Exemple d'utilisation de la fonction markovDecision avec les paramètres layout et circle
 layout = [0, 0, 3, 0, 0, 0, 2, 0, 0, 0, 3, 0, 0, 1, 0]
 
+"""
 # Résolution du problème avec différents modes de jeu
 result_false = markovDecision(layout, circle=False)
 print("\nWin as soon as land on or overstep the final square")
@@ -68,3 +69,4 @@ print(result_false)
 result_true = markovDecision(layout, circle=True)
 print("\nStopping on the square to win")
 print(result_true)
+"""
\ No newline at end of file

plot.py

+import matplotlib.pyplot as plt
+from validation import validation
+import numpy as np
+
+# Example layout and circle settings
+layout = [0, 0, 3, 0, 2, 0, 2, 0, 0, 0, 3, 0, 0, 1, 0]
+circle = False
+
+# Create an instance of validation
+validation_instance = validation(layout, circle)
+
+
+# Plotting function for strategy comparison
+def plot_strategy_comparison(num_games=1000):
+    strategy_costs = validation_instance.compare_strategies(num_games=num_games)
+
+    # Bar plot for strategy comparison
+    plt.figure(figsize=(10, 6))
+    plt.bar(strategy_costs.keys(), strategy_costs.values(), color=['blue', 'green', 'orange', 'red', 'purple'])
+    plt.xlabel('Strategies')
+    plt.ylabel('Average Cost')
+    plt.title('Comparison of Strategies')
+    plt.savefig('strategy_comparison.png')  # Save the plot
+    plt.show()
+
+# Plotting function for state-based average turns for all strategies on the same plot
+def plot_state_based_turns(save=True):
+    strategies = [validation_instance.optimal_strategy,
+                  validation_instance.safe_strategy,
+                  validation_instance.normal_strategy,
+                  validation_instance.risky_strategy,
+                  validation_instance.random_strategy]
+    strategy_names = ['Optimal', 'SafeDice', 'NormalDice', 'RiskyDice', 'Random']
+    
+    plt.figure(figsize=(12, 6))
+    for strategy, name in zip(strategies, strategy_names):
+        mean_turns = validation_instance.simulate_state(strategy, layout, circle)
+        plt.plot(range(len(mean_turns)), mean_turns, marker='o', linestyle='-', label=name)
+
+    plt.xlabel('State')
+    plt.ylabel('Average Turns')
+    plt.title('Average Turns per State for Different Strategies')
+    plt.grid(True)
+    plt.legend()
+
+    #if save:
+         #plt.savefig('state_based_turns_all_strategies.png')  # Save the plot
+
+    plt.show()
+
+def plot_state_based_comparison(validation_instance, num_games=1000):
+    optimal_turns, empirical_turns = validation_instance.compare_state_based_turns(num_games=num_games)
+
+    # Plotting the state-based average turns comparison
+    plt.figure(figsize=(12, 6))
+
+    # Plot optimal strategy turns
+    plt.plot(range(len(optimal_turns)), optimal_turns, marker='o', linestyle='-', label='ValueIteration')
+
+    # Plot empirical strategy turns
+    plt.plot(range(len(empirical_turns)), empirical_turns, marker='x', linestyle='-', label='Empirical')
+
+    plt.xlabel('State')
+    plt.ylabel('Average Turns')
+    plt.title('Average Turns per State - ValueIteration vs. Empirical')
+    plt.grid(True)
+    plt.legend()
+
+    plt.show()
+
+
+
+
+# Main function to generate and save plots
+if __name__ == '__main__':
+    # Example of strategy comparison plot
+    plot_strategy_comparison(num_games=1000)
+
+    # Example of state-based average turns plot for all strategies on the same plot
+    plot_state_based_turns(save=True)
+
+    plot_state_based_comparison(validation_instance, num_games=1000)
\ No newline at end of file

validation.py

@@ -134,6 +134,53 @@ class validation:
 
     def play_random_strategy(self, n_iterations=10000):
         return self.simulate_game(self.random_strategy, n_iterations)
+    
+    def play_empirical_strategy(self):
+        k = 0  # état initial
+        total_turns = 0
+
+        while k < len(self.layout) - 1:
+            action = self.optimal_strategy[k]  # Utiliser la stratégie empirique pour la simulation
+            action_index = int(action) - 1
+            transition_matrix = self.normal_dice  # Utiliser le dé normal pour la stratégie empirique
+
+            # Aplatir la matrice de transition en une distribution de probabilité 1D
+            flattened_probs = transition_matrix[k]
+            flattened_probs /= np.sum(flattened_probs)  # Normalisation des probabilités
+
+            # Mise à jour de l'état (k) en fonction de la distribution de probabilité aplatie
+            k = np.random.choice(len(self.layout), p=flattened_probs)
+
+            # Mise à jour du nombre de tours en fonction de l'état actuel
+            if self.layout[k] == 3 and action == 2:
+                total_turns += 1 if np.random.uniform(0, 1) < 0.5 else 2
+            elif self.layout[k] == 3 and action == 3:
+                total_turns += 2
+            else:
+                total_turns += 1
+
+        return total_turns
+
+    
+    def compare_empirical_vs_value_iteration(self, num_games=1000):
+        value_iteration_turns = self.simulate_state(self.optimal_strategy, self.layout, self.circle)
+        empirical_turns = self.simulate_state(self.optimal_strategy, self.layout, self.circle, n_iterations=num_games)
+
+        # Calculer la moyenne des tours pour chaque état
+        mean_turns_by_state = {
+            'ValueIteration': value_iteration_turns.tolist(),
+            'Empirical': empirical_turns.tolist()
+        }
+
+        return mean_turns_by_state
+    
+    def compare_state_based_turns(self, num_games=1000):
+        optimal_turns = self.simulate_state(self.optimal_strategy, self.layout, self.circle, n_iterations=num_games)
+        empirical_turns = self.simulate_state(self.optimal_strategy, self.layout, self.circle, n_iterations=num_games)
+
+        return optimal_turns, empirical_turns
+
+
 
     def compare_strategies(self, num_games=1000):
         optimal_cost = self.simulate_game(self.optimal_strategy, n_iterations=num_games)
@@ -157,6 +204,26 @@ circle = False
 validation_instance = validation(layout, circle)
 
 
+# Comparer la stratégie empirique avec la stratégie de value iteration
+turns_by_state = validation_instance.compare_empirical_vs_value_iteration(num_games=1000)
+
+# Imprimer les moyennes des tours pour chaque état
+num_states = len(layout)
+for state in range(num_states - 1):
+    print(f"État {state}:")
+    print(f"   ValueIteration - Tours moyens : {turns_by_state['ValueIteration'][state]:.2f}")
+    print(f"   Empirical - Tours moyens : {turns_by_state['Empirical'][state]:.2f}")
+
+# Exécuter la stratégie empirique une fois
+empirical_strategy_result = validation_instance.play_empirical_strategy()
+print("Coût de la stratégie empirique sur un tour :", empirical_strategy_result)
+
+# Comparer la stratégie empirique avec la stratégie de value iteration sur plusieurs jeux
+comparison_result = validation_instance.compare_empirical_vs_value_iteration(num_games=1000)
+print("Coût moyen de la stratégie de value iteration :", comparison_result['ValueIteration'])
+print("Coût moyen de la stratégie empirique :", comparison_result['Empirical'])
+"""
+
 optimal_cost = validation_instance.play_optimal_strategy(n_iterations=10000)
 print("Optimal Strategy Cost:", optimal_cost)
 
@@ -189,3 +256,5 @@ print("Mean Turns for Risky Dice Strategy:", mean_turns_risky_dice)
 random_dice_strategy = validation_instance.random_strategy
 mean_turns_random_dice = validation_instance.simulate_state(random_dice_strategy, layout, circle, n_iterations=10000)
 print("Mean Turns for Random Dice Strategy:", mean_turns_random_dice)
+"""
+