public MultiResValue(IEqualityComparer <int[]> StateComparer, IEqualityComparer <actionType> ActionComparer, List <actionType> AvailableActions, int[] StartState, params object[] numLevels_minLevel)
: base(StateComparer, ActionComparer, AvailableActions, StartState, numLevels_minLevel)
{
minLevel = numLevels_minLevel.Length >= 2 ? (int)numLevels_minLevel[1] : 0;
models = new ModelBasedValue <int[], actionType> [(int)numLevels_minLevel[0]];
transitions = new StateTransition <int[], actionType> [(int)numLevels_minLevel[0]];
subgoals = new List <Goal <int[], actionType> > [(int)numLevels_minLevel[0]];
availableActions = AvailableActions;
stateComparer = StateComparer;
actionComparer = ActionComparer;
stateTree = new intStateTree();
//stateTree = new taxiStateTree();
//stateTree = new learnedStateTree();
pathFinder = new PathFinder <int[], actionType>(stateComparer);
for (int i = 0; i < models.Length; i++)
{
models[i] = new ModelBasedValue <int[], actionType>(stateComparer, actionComparer, availableActions, StartState)
{
maxUpdates = i == 0 ? (minLevel > 0 ? 20 : 20) : 20,
defaultQ = i == 0 ? 15 : 0,
gamma = i == 0 ? 0.9 : 0.4
};
transitions[i] = new StateTransition <int[], actionType>(null, default(actionType), double.NegativeInfinity, null);
subgoals[i] = new List <Goal <int[], actionType> >();
}
currentGoal = new Goal <int[], actionType>(0, null, default(actionType), null, 0, stateComparer, actionComparer);
}
//Given the Current state, needs to view all of the existing model values for the state
//Will return the probabilities from the t-Table values from all the models as a list
public List <double> t_TableValues(StateTransition <stateType, actionType> StaTran,
Dictionary <int, ActionValue <stateType, actionType> > models)
{
List <double> returnValues = new List <double>();
for (int key = 0; key < models.Count; key++)
{
ModelBasedValue <stateType, actionType> modelsCopy = (ModelBasedValue <stateType, actionType>)models[key].T.GetStateValueTable;
Dictionary <stateType, int> s2Counts = modelsCopy.T.GetStateValueTable(StaTran.oldState, StaTran.action);
double thisS2Counts = 0;
if (s2Counts.ContainsKey(StaTran.newState))
{
thisS2Counts = (double)s2Counts[StaTran.newState];
}
double total = (double)s2Counts.Values.Sum();
if (total == 0)
{
returnValues.Add(0);
}
else
{
returnValues.Add(thisS2Counts / total);
}
}
return(returnValues);
}
public LSValue(IEqualityComparer <stateType> StateComparer, IEqualityComparer <actionType> ActionComparer, List <actionType> AvailableActions, stateType StartState, params object[] parameters)
: base(StateComparer, ActionComparer, AvailableActions, StartState, parameters)
{
stateComparer = StateComparer;
actionComparer = ActionComparer;
availableActions = AvailableActions;
trueModel = new ModelBasedValue <stateType, actionType>(StateComparer, ActionComparer, availableActions, StartState);
trueModel.maxUpdates = 1000;
}
public LinearEgoAlloValue(IEqualityComparer <int[]> StateComparer, IEqualityComparer <int[]> ActionComparer, List <int[]> AvailableActions, int[] StartState, params object[] parameters)
: base(StateComparer, ActionComparer, AvailableActions, StartState, parameters)
{
if (parameters.Length > 0)
{
fullPredictionMode = (bool)parameters[0];
}
if (parameters.Length > 1)
{
updateTerminationStepCount = (int)parameters[1];
}
if (parameters.Length > 2)
{
if ((bool)parameters[2])
{
alloLearner = new ModelBasedValue <int[], int[]>(StateComparer, ActionComparer, AvailableActions, null, true);
}
else
{
alloLearner = new ModelFreeValue <int[], int[]>(StateComparer, ActionComparer, AvailableActions, null, true);
}
}
stateComparer = StateComparer;
actionComparer = ActionComparer;
availableActions = AvailableActions;
//alloModel = new ModelFreeValue<int[], int[]>(StateComparer, ActionComparer, availableActions, null, true);
//{
// defaultQ = 10.3
//};
egoLearner = new ModelFreeValue <int[], int[]>(StateComparer, actionComparer, availableActions, StartState)
{
alpha = 0.9
};
egoPredictionModels = new ModelBasedValue <int[], int[]> [3];
for (int i = 0; i < egoPredictionModels.Length; i++)
{
egoPredictionModels[i] = new ModelBasedValue <int[], int[]>(StateComparer, ActionComparer, availableActions, StartState)
{
gamma = 0
};
}
visitedStates = new Dictionary <int[], int> [4];
for (int i = 0; i < availableActions.Count; i++)
{
visitedStates[i] = new Dictionary <int[], int>(StateComparer);
}
}
public ContextChangeValueGeneric(IEqualityComparer <stateType> StateComparer, IEqualityComparer <actionType> ActionComparer, List <actionType> AvailableActions, stateType StartState, params object[] parameters) : base(StateComparer, ActionComparer, AvailableActions, StartState, parameters)
{
//Constructor
ModelBasedValue <stateType, actionType> first = new ModelBasedValue <stateType, actionType>(StateComparer, ActionComparer, AvailableActions, StartState, parameters);
models = new Dictionary <int, ActionValue <stateType, actionType> >();
models.Add(0, first);
activeModelKey = 0;
modelType = first.GetType();
SC = StateComparer;
AC = ActionComparer;
AA = AvailableActions;
SS = StartState;
parames = parameters;
}
public EgoAlloValue(IEqualityComparer <int[]> StateComparer, IEqualityComparer <int[]> ActionComparer, List <int[]> AvailableActions, int[] StartState, params object[] parameters)
: base(StateComparer, ActionComparer, AvailableActions, StartState, parameters)
{
if (parameters.Length > 0)
{
fullPredictionMode = (bool)parameters[0];
}
stateComparer = StateComparer;
actionComparer = ActionComparer;
availableActions = AvailableActions;
alloModel = new ModelBasedValue <int[], int[]>(StateComparer, ActionComparer, availableActions, StartState, true)
{
defaultQ = 10.3
};
egoModel = new ModelFreeValue <int[], int[]>(StateComparer, actionComparer, availableActions, StartState)
{
alpha = 0.9
};
network = new ActivationNetwork(new BipolarSigmoidFunction(2), 10, 10, 3);
teacher = new ParallelResilientBackpropagationLearning(network);
}
