private Value GetExpectation(IMarkovAction <TState> action, TState state, IValueFunction <TState> previousFunction)
{
var expectation = new Value(0.0, 0.0);
foreach (var nextState in action[state])
{
var probability = action[state, nextState];
if (probability == 0)
{
continue;
}
var totalReward = 0.0;
foreach (var reward in Rewards)
{
totalReward += reward[state, nextState, action];
}
var prevValue = previousFunction[nextState];
var rewardValue = new Value(totalReward + (Discount * prevValue.Mean), prevValue.Variance + (totalReward * totalReward) + (2 * Discount * prevValue.Mean * totalReward));
expectation += probability * rewardValue;
}
return(expectation);
}
public Neuron(IValue input, IValueFunction function, double bias)
{
this.input = input;
this.function = function;
this.bias = bias;
}
public TResult ExecuteFunction <TResult>(IValueFunction <TResult> valueFunction)
{
GetOrCreateContext(valueFunction.GetType());
return(valueFunction.Execute(_wrappedContext));
}
