public int policy_per_state(State main_node)
{
int main_node_index = main_node.state_no;
//get all nodes which are connected to main index_state node
List <int> edgeList = Ad_Graph.get_linked_nodes(main_node_index);
double[] pi = new double[no_action];
foreach (int connected_node_index in edgeList)
{
//State connected_node= Global_nodes.Find(x => x.state_no == connected_node_index); //use a map instead fo linear search for find
State connected_node = Global_nodes[state_map[connected_node_index]];
double connected_node_value = connected_node.state_value;
for (int j = 0; j < no_action; j++)
{
var trans = new Tuple <string, int, string>(main_node_index.ToString(), j + 1, connected_node_index.ToString()); //here so many keys wont be found , i can eithermake others transision as zero
if (trans_prob.ContainsKey(trans))
{
pi[j] += trans_prob[trans] * connected_node_value;
}
}
}
double max = pi[0];
int pos = 0;
for (int i = 0; i < pi.Length; i++)
{
if (pi[i] > max)
{
max = pi[i]; pos = i;
}
}
//Console.WriteLine("pi: {0} {1} {2}", pi[0],pi[1],pi[2]);
return(pos + 1);
}
public void state_flow(int start_state)
{
if (start_state == 0)
{
writer.WriteLine(string.Format("{0}", "0"));
return;
}
List <int> edgeList = Ad_Graph.get_linked_nodes(start_state);
int best_action = mdp_model[start_state.ToString()].Item2;
Double prob = 0;
Double max_prob = 0;
int state_max_prob = 0;
//Console.WriteLine(start_state + " " + best_action);
var row = string.Format("{0},{1}", start_state, best_action);
writer.WriteLine(row);
foreach (int connected_node_index in edgeList) //calculate utility for each node
{
var trans = new Tuple <string, int, string>(start_state.ToString(), best_action, connected_node_index.ToString()); //here so many kets wont be found , i can eithermake others transision as zero
if (trans_prob.ContainsKey(trans))
{
prob = trans_prob[trans];
if (prob > max_prob)
{
max_prob = prob;
state_max_prob = connected_node_index;
}
}
}
state_flow(state_max_prob);
}