Skip to content
Extraits de code Groupes Projets
Valider 399700c8 rédigé par Adrien Payen's avatar Adrien Payen
Parcourir les fichiers

commit ML

parent 7707bd45
Aucune branche associée trouvée
Aucune étiquette associée trouvée
Aucune requête de fusion associée trouvée
import numpy as np
import random as rd
import matplotlib.pyplot as plt
from tmc import TransitionMatrixCalculator as tmc
from markovDecision import markovDecision as mD
from validation import Validation
def plot_results(validation_instance):
results_markov = validation_instance.simulate_game('markov')
results_safe = validation_instance.simulate_game([1]*15)
results_normal = validation_instance.simulate_game([2]*15)
results_risky = validation_instance.simulate_game([3]*15)
results_random = validation_instance.simulate_game(np.random.randint(1, 4, size=15))
plt.figure(figsize=(12, 8))
plt.plot(range(len(validation_instance.layouts)), results_markov, label='Markov')
plt.plot(range(len(validation_instance.layouts)), results_safe, label='SafeDice')
plt.plot(range(len(validation_instance.layouts)), results_normal, label='NormalDice')
plt.plot(range(len(validation_instance.layouts)), results_risky, label='RiskyDice')
plt.plot(range(len(validation_instance.layouts)), results_random, label='Random')
plt.xticks(range(len(validation_instance.layouts)), range(len(validation_instance.layouts)))
plt.xlabel('Layout Number', fontsize=13)
plt.ylabel('Average Number of Turns', fontsize=13)
plt.legend(loc='upper left', bbox_to_anchor=(1, 1), ncol=1)
plt.show()
# Example usage
layouts = [
[0, 0, 3, 0, 0, 0, 2, 0, 0, 0, 3, 0, 0, 1, 0],
# Add more layouts as needed
]
validation_instance = Validation(layouts, circle=False, n_iterations=10000)
plot_results(validation_instance)
\ No newline at end of file
......@@ -217,7 +217,7 @@ class TransitionMatrixCalculator:
# create a function that test the transition matrix for different layout each time with one trap of each sort
def tst_transition_matrix(self):
# create a list of 100 different layouts
layouts = self.generate_arrays(1000)
layouts = self.generate_arrays(100)
for array in layouts:
print(array)
self.compute_transition_matrix(array, False)
......
import random as rd
import numpy as np
import matplotlib.pyplot as plt
from tmc import TransitionMatrixCalculator as tmc
from markovDecision import markovDecision as mD
class EmpiricalComparision :
def __init__(self) :
return
def simulation(strategy, layout : list, circle, nIter : int) :
tmc_instance = tmc()
safe_dice = tmc_instance._compute_safe_matrix(layout, circle)
normal_dice = tmc_instance._compute_normal_matrix(layout, circle)
risky_dice = tmc_instance._compute_risky_matrix(layout, circle)
matrices_transition = [safe_dice, normal_dice, risky_dice]
nTurns = []
turns = 0
for _ in range(nIter) :
turns = 0
k = 0
while k < len(layout)-1 :
action = strategy[k]
transitionMatrix = matrices_transition[int(action -1)]
k = np.rd.choice(len(layout), p = transitionMatrix[k])
if layout[k] == 3 and action == 2 :
turns +=1 if np.rd.uniform(0,1) < 0.5 else 2
elif layout[k] == 3 and action == 3 :
turns += 2
else :
turns += 1
nTurns.append(turns)
return np.mean(nTurns)
def plot(layouts : list, circle, nIter : int) :
Markov = []
Safe = []
Normal = []
Risky = []
Random = []
for layout in layouts :
expec, policy = mD(layout, circle)
# Simulate the game
return
layout = [0,0,3,0,0,0,2,0,0,0,3,0,0,1,0]
results(layout, False, 1000000)
results(layout, True, 1000000)
\ No newline at end of file
from tmc import TransitionMatrixCalculator
class Validation:
def __init__(self, layout, circle=False):
self.layout = layout
self.circle = circle
self.tmc_instance = TransitionMatrixCalculator()
def simulate_game(self, strategy='optimal', num_games=1000):
total_turns = 0
for _ in range(num_games):
if strategy == 'optimal':
turns = self.play_optimal_strategy()
elif strategy == 'dice1':
turns = self.play_dice_strategy(1)
elif strategy == 'dice2':
turns = self.play_dice_strategy(2)
elif strategy == 'dice3':
turns = self.play_dice_strategy(3)
elif strategy == 'random':
turns = self.play_random_strategy()
total_turns += turns
average_turns = total_turns / num_games
return average_turns
def play_optimal_strategy(self):
# Implement the optimal strategy using value iteration results
# Use TransitionMatrixCalculator to compute transitions and make decisions
# calculer la stratégie optimale pour ou un tour
pass
def play_dice_strategy(self, dice):
# Implement a strategy where only one type of dice is used (1, 2, or 3)
pass
def play_random_strategy(self):
# Implement a purely random strategy
pass
def compare_strategies(self, num_games=1000):
optimal_cost = self.simulate_game(strategy='optimal', num_games=num_games)
dice1_cost = self.simulate_game(strategy='dice1', num_games=num_games)
dice2_cost = self.simulate_game(strategy='dice2', num_games=num_games)
dice3_cost = self.simulate_game(strategy='dice3', num_games=num_games)
random_cost = self.simulate_game(strategy='random', num_games=num_games)
return {
'optimal': optimal_cost,
'dice1': dice1_cost,
'dice2': dice2_cost,
'dice3': dice3_cost,
'random': random_cost
}
# Example usage
layout = [0, 0, 3, 0, 0, 0, 2, 0, 0, 0, 3, 0, 0, 1, 0]
validation = Validation(layout, circle=False)
results = validation.compare_strategies(num_games=10000)
print("Average Costs:")
for strategy, cost in results.items():
print(f"{strategy}: {cost}")
import numpy as np
from tmc import TransitionMatrixCalculator
import random
import matplotlib.pyplot as plt
from markovDecision import markovDecision
class Validation:
def __init__(self, layout, circle=False):
self.layout = layout
self.circle = circle
self.tmc_instance = TransitionMatrixCalculator()
def simulate_game(self, strategy='optimal', num_games=1000):
total_turns = 0
for _ in range(num_games):
if strategy == 'optimal':
turns = self.play_optimal_strategy()
elif strategy == 'dice1':
turns = self.play_dice_strategy(1)
elif strategy == 'dice2':
turns = self.play_dice_strategy(2)
elif strategy == 'dice3':
turns = self.play_dice_strategy(3)
elif strategy == 'random':
turns = self.play_random_strategy()
total_turns += turns
average_turns = total_turns / num_games
return average_turns
def play_optimal_strategy(self):
_, optimal_policy = markovDecision(self.layout, self.circle)
return self.empirical_results(optimal_policy.astype(int))
def play_dice_strategy(self, dice):
policy = np.ones(len(self.layout), dtype=int) * dice
return self.empirical_results(policy)
def play_random_strategy(self):
policy = np.zeros(len(self.layout), dtype=int)
for i in range(len(policy) - 1):
policy[i] = random.choice([1, 2, 3])
return self.empirical_results(policy)
def empirical_results(self, policy):
avgnTurnsPlayed = 0
nSimul = 10000
for _ in range(nSimul):
nTurns = self.playOneGame(policy)
avgnTurnsPlayed += nTurns
return avgnTurnsPlayed / nSimul
def playOneGame(self, policy):
nSquares = len(self.layout)
nTurns = 0
curPos = 0
jail = False
while curPos < nSquares - 1:
newPos, jail = self.playOneTurn(policy[curPos], curPos)
curPos = newPos
nTurns += 1
return nTurns
def playOneTurn(self, diceChoice, curPos):
nSquares = len(self.layout)
if curPos == nSquares - 1:
return nSquares - 1, False
if jail :
return curPos, False
listDiceResults = [i for i in range(diceChoice + 1)]
result = random.choice(listDiceResults)
if curPos == 2 and result != 0:
slowLane = random.choice([0, 1])
if slowLane:
newPos = curPos + result
else:
newPos = curPos + result + 7
elif ((curPos == 9 and result != 0) or ((curPos in [7, 8, 9]) and (curPos + result >= 10))):
newPos = curPos + result + 4
else:
newPos = curPos + result
if newPos > nSquares - 1:
if self.circle:
newPos -= nSquares
else:
return nSquares - 1, True
newSquare = self.layout[newPos]
if diceChoice == 1:
return newPos, False
elif diceChoice == 2:
newSquare = random.choice([0, newSquare])
if newSquare == 0:
return newPos, False
elif newSquare == 1:
return 0, False
elif newSquare == 2:
if newPos - 3 < 0:
return 0, False
return newPos - 3, False
elif newSquare == 3:
return newPos, True
elif newSquare == 4:
newSquare = random.choice([1, 2, 3])
if newSquare == 1:
return 0, False
elif newSquare == 2:
if newPos - 3 < 0:
return 0, False
return newPos - 3, False
elif newSquare == 3:
return newPos, True
def compare_strategies(self, num_games=1000):
optimal_cost = self.simulate_game(strategy='optimal', num_games=num_games)
dice1_cost = self.simulate_game(strategy='dice1', num_games=num_games)
dice2_cost = self.simulate_game(strategy='dice2', num_games=num_games)
dice3_cost = self.simulate_game(strategy='dice3', num_games=num_games)
random_cost = self.simulate_game(strategy='random', num_games=num_games)
return {
'optimal': optimal_cost,
'dice1': dice1_cost,
'dice2': dice2_cost,
'dice3': dice3_cost,
'random': random_cost
}
# Example usage
layout = [0, 0, 3, 0, 0, 0, 2, 0, 0, 0, 3, 0, 0, 1, 0]
validation = Validation(layout, circle=False)
results = validation.compare_strategies(num_games=10000)
print("Average Costs:")
for strategy, cost in results.items():
print(f"{strategy}: {cost}")
import random as rd
import numpy as np
import matplotlib.pyplot as plt
from tmc import TransitionMatrixCalculator as tmc
from markovDecision import markovDecision as mD
class EmpiricalComparision :
def __init__(self) :
return
def simulation(strategy, layout : list, circle, nIter : int) :
tmc_instance = tmc()
safe_dice = tmc_instance._compute_safe_matrix(layout, circle)
normal_dice = tmc_instance._compute_normal_matrix(layout, circle)
risky_dice = tmc_instance._compute_risky_matrix(layout, circle)
matrices_transition = [safe_dice, normal_dice, risky_dice]
nTurns = []
turns = 0
for _ in range(nIter) :
turns = 0
k = 0
while k < len(layout)-1 :
action = strategy[k]
transitionMatrix = matrices_transition[int(action -1)]
k = np.rd.choice(len(layout), p = transitionMatrix[k])
if layout[k] == 3 and action == 2 :
turns +=1 if np.rd.uniform(0,1) < 0.5 else 2
elif layout[k] == 3 and action == 3 :
turns += 2
else :
turns += 1
nTurns.append(turns)
return np.mean(nTurns)
def plot(layouts : list, circle, nIter : int) :
Markov = []
Safe = []
Normal = []
Risky = []
Random = []
for layout in layouts :
expec, policy = mD(layout, circle)
# Simulate the game
return
layout = [0,0,3,0,0,0,2,0,0,0,3,0,0,1,0]
results(layout, False, 1000000)
results(layout, True, 1000000)
\ No newline at end of file
0% Chargement en cours ou .
You are about to add 0 people to the discussion. Proceed with caution.
Terminez d'abord l'édition de ce message.
Veuillez vous inscrire ou vous pour commenter