認知情報解析演習a Monte Carlo 01

north=c(1:3,15,1:10)
east=2:15;east[ c(3,7,11)]=c(3,7,11)
south=c(5:15,12:14)
west=c(15,1:13);west[ c(4,8,12)]=c(4,8,12)
trM=cbind(north,east,south,west)

r=-1;P=rep(0.25,4);V = rep(0,14)max.iter = 10000;
state.count=rep(0,15)
for (i.iter in 1:max.iter){
  state = sample(1:14,1)
  state.seq = state
  while(state!=15){
    action = sample(1:4,1,prob = P)
    state.seq = c(state.seq,trM[state,action])
    state = trM[state,action]  
  }
  uniq.seq = unique(state.seq)
  for (i.uniq in 1:(length(uniq.seq)-1)){
    first.visit = which(state.seq == uniq.seq[i.uniq])[1]
    V[uniq.seq[i.uniq]] = V[uniq.seq[i.uniq]] + r*(length(state.seq)-first.visit-1)
  }
  state.count[uniq.seq] = state.count[uniq.seq] + 1
}
V = matrix(c(0,V/state.count[1:14],0),nrow=4)