def __init__(self, question, testDict):
super(EpsilonGreedyTest, self).__init__(question, testDict)
self.discount = float(testDict['discount'])
self.grid = gridworld.Gridworld(parseGrid(testDict['grid']))
if 'noise' in testDict: self.grid.setNoise(float(testDict['noise']))
if 'livingReward' in testDict: self.grid.setLivingReward(float(testDict['livingReward']))
self.grid = gridworld.Gridworld(parseGrid(testDict['grid']))
self.env = gridworld.GridworldEnvironment(self.grid)
self.epsilon = float(testDict['epsilon'])
self.learningRate = float(testDict['learningRate'])
self.numExperiences = int(testDict['numExperiences'])
self.numIterations = int(testDict['iterations'])
self.opts = {'actionFn': self.env.getPossibleActions, 'epsilon': self.epsilon, 'gamma': self.discount, 'alpha': self.learningRate}
def __init__(self, question, testDict):
super(GridPolicyTest, self).__init__(question, testDict)
# Function in module in analysis that returns (discount, noise)
self.parameterFn = testDict['parameterFn']
self.question2 = testDict.get('question2', 'false').lower() == 'true'
# GridWorld specification
# _ is empty space
# numbers are terminal states with that value
# # is a wall
# S is a start state
#
self.gridText = testDict['grid']
self.grid = gridworld.Gridworld(parseGrid(testDict['grid']))
self.gridName = testDict['gridName']
# Policy specification
# _ policy choice not checked
# N, E, S, W policy action must be north, east, south, west
#
self.policy = parseGrid(testDict['policy'])
# State the most probable path must visit
# (x,y) for a particular location; (0,0) is bottom left
# terminal for the terminal state
self.pathVisits = testDict.get('pathVisits', None)
# State the most probable path must not visit
# (x,y) for a particular location; (0,0) is bottom left
# terminal for the terminal state
self.pathNotVisits = testDict.get('pathNotVisits', None)
def __init__(self, question, testDict):
super(QLearningTest, self).__init__(question, testDict)
self.discount = float(testDict['discount'])
self.grid = gridworld.Gridworld(parseGrid(testDict['grid']))
if 'noise' in testDict: self.grid.setNoise(float(testDict['noise']))
if 'livingReward' in testDict: self.grid.setLivingReward(float(testDict['livingReward']))
self.grid = gridworld.Gridworld(parseGrid(testDict['grid']))
self.env = gridworld.GridworldEnvironment(self.grid)
self.epsilon = float(testDict['epsilon'])
self.learningRate = float(testDict['learningRate'])
self.opts = {'actionFn': self.env.getPossibleActions, 'epsilon': self.epsilon, 'gamma': self.discount, 'alpha': self.learningRate}
numExperiences = int(testDict['numExperiences'])
maxPreExperiences = 10
self.numsExperiencesForDisplay = range(min(numExperiences, maxPreExperiences))
self.testOutFile = testDict['test_out_file']
if maxPreExperiences < numExperiences:
self.numsExperiencesForDisplay.append(numExperiences)
def __init__(self, question, testDict):
super(ValueIterationTest, self).__init__(question, testDict)
self.discount = float(testDict['discount'])
self.grid = gridworld.Gridworld(parseGrid(testDict['grid']))
iterations = int(testDict['valueIterations'])
if 'noise' in testDict: self.grid.setNoise(float(testDict['noise']))
if 'livingReward' in testDict: self.grid.setLivingReward(float(testDict['livingReward']))
maxPreIterations = 10
self.numsIterationsForDisplay = range(min(iterations, maxPreIterations))
self.testOutFile = testDict['test_out_file']
if maxPreIterations < iterations:
self.numsIterationsForDisplay.append(iterations)