a_matrix = np.zeros((len(state_to_index), len(action_state_to_index)), dtype=np.float_) for index2, val in enumerate(action_state_to_index): state=val[0] state_index=state_to_index.index(val[0]) action=val[1] next_states_list=actions_dic[val[0]][val[1]] if action=="NONE": a_matrix[state_index][index2]+=1.0 else: for x in range(0, len(next_states_list)-1, 2): a_matrix[state_index][index2] += next_states_list[x+1] next_state_index=state_to_index.index(next_states_list[x]) a_matrix[next_state_index][index2] -= next_states_list[x+1]