# For saving toplist
objective = -1
top_UI = params.top * [] #([([],[],policies,100)])*params.top
top_UI_summary = [] # []*params.top #np.zeros((params.top,3))
multi_obj_topUI = [[], [], []]
##########################################
# TODO: READ UI from the text file uis/remote.txt
UI = np.array([[0, 1, 1, 1], [0, 0, 1, 1], [0, 1, 0, 1], [0, 1, 1, 0]])
num_action = params.num_states - 1 # Number of unique buttons
#############################
# Call RL code
objective, top_UI, best_actions, best_path, klm_value = rl_optimizer(
UI, num_action, top_UI, params, batch_num, logging)
if objective == -1: # If not learned, skip all button combinations
print "Skipping this UI"
###################
# Write to file
output_file = open('best_path.txt', 'w')
#output_file.writelines(tags[int(best_path[i])]+"\n")
for i in range(np.size(best_path)):
output_file.writelines(tags[int(best_path[i])] + "\n")
#print tags[int(best_path[i])]
output_file.writelines(str(klm_value))
output_file.close()
tags.append(item_info[0])
x_coord[idx] = float(item_info[1])
y_coord[idx] = float(item_info[2])
widths[idx] = float(item_info[3])
heights[idx] = float(item_info[4])
idx += 1
###################
# Initialize parameters
params = initializeParams(x_coord, y_coord, widths, heights)
num_action = params.num_states-1
###################
# Call RL code
best_path, klm_value = rl_optimizer(num_action, params)
###################
# Write to file
output_file = open('best_path.txt', 'w')
for i in range(np.size(best_path)):
output_file.writelines(tags[int(best_path[i])]+"\n")
output_file.writelines(str(klm_value))
output_file.close()
# Penalty index for each action. The penalty values in params.penalties
actions_penalty = [[1, 2, 3, 4], [5, 6, 7, 8], [9, 10, 11, 12]]
print "Sensor errors", params.sensor_errors
print "Confusion errors", params.confusion_error
#############################
## Learning policies and calculating KLM esitemate
modality_table_total = np.array([0, 0, 0])
klm_total = 0
# If sensor errors are used the number of iterations should be more than 1 to average the effect of errors.
num_iters = 5
for j in range(num_iters):
modality_table, klm = rl_optimizer(UIs, actionmatrix, actions_in_uis,
actions_penalty, num_commands, params)
modality_table_total += modality_table
# KLM estimates
klm_total += klm
print "Tactile commands used ", modality_table_total[0], "times."
print "Gesture commands used ", modality_table_total[1], "times."
print "Speech commands used ", modality_table_total[2], "times."
# Average KLM estimate over all iterations
print "KLM estimate:", klm_total / num_iters
random_search_idx = random.randint(0, len(button_matrices)-1)
while random_search_idx in search_past_idxes:
random_search_idx = random.randint(0, len(button_matrices)-1)
search_past_idxes.append(random_search_idx)
buttonmatrix = button_matrices[random_search_idx]
else:
if search_iter == len(button_matrices):
break
buttonmatrix = button_matrices[search_iter]
##########################
# Train RL agent
##########################
start_time_rl = time.time()
top_buttons, objective = rl_optimizer(UI, buttonmatrix, num_buttons, top_buttons, params, batch_num, logging)
total_time_rl += (time.time()-start_time_rl)
if objective == -1: # If not learned, skip all button combinations
print "Skipping this UI"
break
search_iter += 1
for k in range(len(top_buttons)):
top_UI.append(top_buttons[k])
top_UI = sorted(top_UI, key=op.itemgetter(3))[:params.the_top]
[0,4,5,6,7,8,0,10,11,12],
[0,4,5,6,7,8,9,0,11,12],
[0,4,5,6,7,8,9,10,0,12],
[0,4,5,6,7,8,9,10,11,0]])
# Allowed actions in eac state
actionmatrix = np.array([[1],[1,2,3,5,6,7,8,9,10,11,12],
[1,2,3,4,6,7,8,9,10,11,12],
[1,2,3,4,5,7,8,9,10,11,12],
[1,2,3,4,5,6,8,9,10,11,12],
[1,2,3,4,5,6,7,9,10,11,12],
[1,2,3,4,5,6,7,8,10,11,12],
[1,2,3,4,5,6,7,8,9,11,12],
[1,2,3,4,5,6,7,8,9,10,12],
[1,2,3,4,5,6,7,8,9,10,11]])
num_buttons = 12 # Number of unique buttons
#All Uis
UIs = [UI1, UI2]
actions_in_uis = [[1,2,3],[4,5,6,7,8,9,10,11,12]] # [[1,2,3]] #
actions_penalty = [[],[2,3],[1,4,5,6,7,8,9,10,11,12]]
#############################
# Call RL code
klm = rl_optimizer(UIs, actionmatrix, actions_in_uis, actions_penalty, num_buttons, params)
print "KLM average:", klm