This program has been disqualified.
| Author | Deltcho |
| Submission date | 2012-01-23 04:39:05.597644 |
| Rating | 5000 |
| Matches played | 0 |
| Win rate | 0 |
import math as np
import random
global val_set, item_occurance, LTM_limit, STM_limit, memory_array, weighted_memory_array, memory_array_index, win_loss_array,win_loss_weight, WLW_STM, MAI_STM
global STM_Halflife, LTM_Halflife, mindset, MA_STM, WMA_STM, WLA_STM, streak, previous_input
global STM_w_bias, STM_w_laystats, STM_w_prediction, LTM_w_prediction, LTM_w_bias, STM_model, learning
if input == "":
output = random.choice(["R","P","S"])
winning = []
STM_w_bias = 0.40
STM_w_laystats = 0.15
STM_w_prediction = 0.45
LTM_w_prediction = 0.65
LTM_w_bias = 0.35
memory_array_index = 0
MAI_STM = 0
streak = 5
previous_input = ''
STM_Halflife = 5
LTM_Halflife = 25
LTM_limit = 50
STM_limit = 7
STM_model = 3
learning = 1
mindset = []
memory_array = []
weighted_memory_array = []
win_loss_array = []
win_loss_weight = []
MA_STM = []
WMA_STM = []
WLA_STM = []
WLW_STM = []
for i in range(STM_limit):
MA_STM.append(0)
WMA_STM.append(0)
WLA_STM.append(0)
WLW_STM.append(0)
for i in range(LTM_limit):
memory_array.append(0)
weighted_memory_array.append(0)
win_loss_array.append(0)
win_loss_weight.append(0)
val_set = ['R','P','S']
item_occurance = []
for i in range(len(val_set)):
item_occurance.append(0)
def weighted_choice(a):
random1 = random.uniform(0,1)
if random1 <= a[0]:
return 'R'
elif random1 > a[0] and random1 <= (a[0]+a[1]):
return 'P'
elif random1 > (a[0]+a[1]):
return 'S'
def average(values):
return sum(values, 0.0) / len(values)
def sum_moo(values):
return sum(values, 0.0)
def STM (inp,output):
global val_set, item_occurance, STM_limit, WLW_STM, MAI_STM
global STM_Halflife, mindset, MA_STM, WMA_STM, WLA_STM, streak, previous_input
if MAI_STM == (STM_limit):
LTM_content = LTM(MA_STM[0][0],MA_STM[0][1], WLA_STM[0])
MA_STM = MA_STM[1:]
MA_STM.append([inp,output])
WLA_STM = WLA_STM[1:]
WLA_STM.append(RPS(inp,output))
for (i) in range(MAI_STM):
WMA_STM[i] = salience((MAI_STM-(1+i)),'STM')
previous_input = MA_STM[MAI_STM-1][1]
elif MAI_STM < (STM_limit):
MA_STM[MAI_STM] = [inp,output]
previous_input = MA_STM[MAI_STM][0]
WLA_STM[MAI_STM] = RPS(inp,output)
for (i) in range(MAI_STM):
WMA_STM[i] = salience((MAI_STM-1-(i)),'STM')
MAI_STM += 1
for i in range(len(val_set)):
if inp == val_set[i]:
item_occurance[i] += 1
def LTM(inp,output, WLA_value):
global val_set, item_occurance, LTM_limit, memory_array, weighted_memory_array, memory_array_index, win_loss_array,win_loss_weight, streak
if inp and output == 0:
return long_term_memory
else:
if memory_array_index == (LTM_limit):
memory_array = memory_array[1:]
memory_array.append([inp,output])
win_loss_array = win_loss_array[1:]
win_loss_array.append(WLA_value)
elif memory_array_index < (LTM_limit):
win_loss_array[memory_array_index] = WLA_value
memory_array[memory_array_index] = [inp,output]
for (i) in range(memory_array_index):
weighted_memory_array[i] = salience((memory_array_index-(i+1)),'LTM')
memory_array_index += 1
for i in range(len(val_set)):
if inp == val_set[i]:
item_occurance[i] += 1
def salience(memory_position, memory_type):
global STM_Halflife, LTM_Halflife, STM_limit, STM_model
if memory_type == 'STM':
if STM_model == 1:
new_salience = 100*(np.e)**((np.log(0.5)/STM_Halflife)*memory_position)
elif STM_model == 2:
if memory_position <= (int(STM_limit/2)*1+1):
new_salience = 100*(np.e)**((np.log(0.5)/STM_Halflife)*((int(STM_limit/2)*1+1)-memory_position))
else:
new_salience = 100*(np.e)**((np.log(0.5)/STM_Halflife)*(memory_position-(int(STM_limit/2)*1+1)))
elif STM_model == 3:
if memory_position <= (int(STM_limit/2)*1+1):
new_salience = 100*(np.e)**((np.log(0.5)/STM_Halflife)*(memory_position))
else:
new_salience = 100*(np.e)**((np.log(0.5)/STM_Halflife)*(STM_limit-1-memory_position))
elif memory_type == 'LTM':
new_salience = 100*(np.e)**((np.log(0.5)/LTM_Halflife)*memory_position)
return new_salience
def win_loss_salience(MA, MAI, WLA, WLW, f_type):
global streak
if f_type == 'pattern':
for (i) in range(MAI):
WLW[i] = 0
for i in range(MAI):
if (MAI-i) > streak:
for x in range(streak):
WLW[(MAI-1-i-x)] = average(WLA[(MAI-i-streak):(MAI-i)])
else:
for x in range(streak):
WLW[(MAI-1-i-x)] = average(WLA[:(MAI-i)])
w_l_pattern = WLW[:]
return w_l_pattern
if f_type =='bias':
for (i) in range(MAI):
WLW[i] = 0
for i in range(MAI):
if (MAI-i-1) > streak:
for x in range(streak):
WLW[(MAI-1-i-x)] += average(WLA[(MAI-i-streak):(MAI-i)])
else:
for x in range(streak):
WLW[(MAI-1-i-x)] += average(WLA[:(MAI-i)])
for (i) in range(MAI):
WLW[i] = WLW[i]/streak
w_l_bias = WLW[:]
return w_l_bias
def chunks(l, n):
return [l[i:i+n] for i in range(0, len(l), n)]
def prediction(MA, MAI, WMA, last_input):
global val_set
pattern = [[0,0],[0,0]]
val_set_cnt = []
val_set_weight = []
val_set_weight_play = []
for i in range(len(val_set)):
val_set_cnt.append(0)
val_set_weight.append(0)
val_set_weight_play.append(0)
for i in range((MAI-1)):
pattern = [MA[i][0], MA[i+1][0]]
if pattern[0] == last_input:
for x in range(len(val_set)):
if pattern[1] == val_set[x]:
val_set_cnt[x] += 1.0
val_set_weight[x] += WMA[i+1]
if sum_moo(val_set_weight) > 0:
blah_weight = sum_moo(val_set_weight)
for i in range(len(val_set)):
val_set_weight[i] = val_set_weight[i] / blah_weight
for i in range(len(val_set)):
if i > 0:
val_set_weight_play[i] = val_set_weight[(i-1)]
elif i == 0:
val_set_weight_play[i] = val_set_weight[(len(val_set)-1)]
else:
for x in range(len(val_set)):
val_set_weight_play.append(1.0 / float(len(val_set)))
return val_set_weight_play
def superstition(MA, MAI):
global val_set, streak
val_set_cnt = []
val_set_cnt_play = []
for a in range(len(val_set)):
val_set_cnt.append(0)
val_set_cnt_play.append(0)
if MAI > streak:
for i in range(MAI):
for x in range(len(val_set)):
if MA[i][0] == val_set[x]:
val_set_cnt[x] += 1.0
blah_cnt = sum_moo(val_set_cnt)
for i in range(len(val_set)):
val_set_cnt[i] = (1 - (val_set_cnt[i] / blah_cnt))
blah_cnt = sum_moo(val_set_cnt)
for i in range(len(val_set)):
val_set_cnt[i] = (1 - (val_set_cnt[i] / blah_cnt)) / blah_cnt
for i in range(len(val_set)):
if i > 0:
val_set_cnt_play[i] = val_set_cnt[(i-1)]
elif i == 0:
val_set_cnt_play[i] = val_set_cnt[(len(val_set)-1)]
else:
for x in range(len(val_set)):
val_set_cnt_play.append(1.0 / len(val_set))
return val_set_cnt_play
def RPS (inp,output):
if inp == 'R':
if output == 'R':
val = 0.5
elif output == 'P':
val = 1.0
elif output == 'S':
val = 0.0
elif inp == 'P':
if output == 'R':
val = 0.0
elif output == 'P':
val = 0.5
elif output == 'S':
val = 1.0
elif inp == 'S':
if output == 'R':
val = 1.0
elif output == 'P':
val = 0.0
elif output == 'S':
val = 0.5
return val
def pattern_recognition(MA, MAI):
global val_set, streak, MA_STM
pattern = []
val_set_cnt = []
val_set_weight = []
val_set_weight_play = []
pattern_input = []
pattern = []
for i in range(len(val_set)):
val_set_cnt.append(0)
val_set_weight.append(0)
val_set_weight_play.append(0)
for moo in range(STM_limit):
pattern_input.append(MA_STM[moo][0])
for i in range((MAI-streak)):
for x in range(streak):
pattern.append([MA[i+x][0]])
if pattern == pattern_input:
for y in range(len(val_set)):
if MA[i+x+1][0] == val_set[y]:
val_set_cnt[y] += 1.0
pattern = []
blah_weight = sum_moo(val_set_weight)
if sum_moo(val_set_weight) > 0:
for i in range(len(val_set)):
val_set_weight[i] = val_set_weight[i] / blah_weight
for i in range(len(val_set)):
if i > 0:
val_set_weight_play[i] = val_set_weight[(i-1)]
elif i == 0:
val_set_weight_play[i] = val_set_weight[(len(val_set)-1)]
else:
for x in range(len(val_set)):
val_set_weight_play.append(1.0 / float(len(val_set)))
return val_set_weight_play
def decision():
global val_set, item_occurance, LTM_limit, STM_limit, memory_array, weighted_memory_array, memory_array_index, win_loss_array,win_loss_weight, WLW_STM, MAI_STM
global STM_Halflife, LTM_Halflife, mindset, MA_STM, WMA_STM, WLA_STM, streak, previous_input
global STM_output, LTM_output
global STM_w_bias, STM_w_laystats, STM_w_prediction, LTM_w_bias,LTM_w_prediction, STM_w_pat
if MAI_STM < streak:
output = weighted_choice([0.333,0.333,0.333])
else:
val_set_cnt = []
val_set_weight = []
val_set_bias = []
STM_weights = []
LTM_weights = []
for i in range(len(val_set)):
val_set_cnt.append(0)
val_set_weight.append(0)
val_set_bias.append(0)
STM_weights.append(0)
LTM_weights.append(0)
STM_prediction = prediction(MA_STM, MAI_STM, WMA_STM,previous_input)
#STM_laystats = superstition(MA_STM,MAI_STM)
STM_bias_weight = win_loss_salience(MA_STM,MAI_STM, WLA_STM, WLW_STM,'bias')
for i in range(MAI_STM):
for x in range(len(val_set)):
if MA_STM[i][1] == val_set[x]:
val_set_weight[x] += STM_bias_weight[i]
val_set_cnt[x] += 1.0
if sum_moo(val_set_cnt) > 0:
blah_cnt = sum_moo(val_set_cnt)
for i in range(len(val_set)):
val_set_weight[i] = val_set_weight[i] / blah_cnt
blah_weight = sum_moo(val_set_weight)
for i in range(len(val_set)):
if blah_weight > 0:
val_set_bias[i] = val_set_weight[i] / blah_weight
else:
for x in range(len(val_set)):
val_set_bias[x] = 1.0 / len(val_set)
else:
for x in range(len(val_set)):
val_set_bias[x] = 1.0 / len(val_set)
for y in range(len(val_set)):
#STM_weights[y] = (STM_w_bias*val_set_bias[y]/1.0 + STM_w_laystats*STM_laystats[y]/1.0 + STM_w_prediction*STM_prediction[y]/1.0 )/ (STM_w_bias/1.0 + STM_w_laystats/1.0 + STM_w_prediction/1.0)
STM_weights[y] = (STM_w_bias*val_set_bias[y]/1.0 + STM_w_prediction*STM_prediction[y]/1.0 )/ (STM_w_bias/1.0 + STM_w_prediction/1.0)
#STM_output = [val_set_bias, STM_laystats, STM_prediction]
STM_output = [val_set_bias, STM_prediction]
blah_stm = sum_moo(STM_weights)
for i in range(len(val_set)):
STM_weights[i] = (STM_weights[i]/blah_stm)
output = weighted_choice(STM_weights)
if memory_array_index >= streak:
val_set_cnt = []
val_set_weight = []
val_set_bias = []
aggregate_weights = []
for i in range(len(val_set)):
val_set_cnt.append(0)
val_set_weight.append(0)
val_set_bias.append(0)
aggregate_weights.append(0)
LTM_prediction = prediction(memory_array,memory_array_index,weighted_memory_array,previous_input)
LTM_bias_weight = win_loss_salience(memory_array, memory_array_index, win_loss_array, win_loss_weight,'bias')
for i in range(memory_array_index):
for x in range(len(val_set)):
if memory_array[i][1] == val_set[x]:
val_set_weight[x] += LTM_bias_weight[i]
val_set_cnt[x] += 1.0
blah_cnt = sum_moo(val_set_cnt)
if sum_moo(val_set_cnt) > 0:
for i in range(len(val_set)):
val_set_weight[i] = val_set_weight[i] / blah_cnt
if sum_moo(val_set_weight) > 0:
blah_weight = sum_moo(val_set_weight)
for i in range(len(val_set)):
val_set_bias[i] = val_set_weight[i] / blah_weight
else:
for x in range(len(val_set)):
val_set_bias[x] = 1.0 / len(val_set)
else:
for x in range(len(val_set)):
val_set_bias[x] = 1.0 / len(val_set)
for y in range(len(val_set)):
LTM_weights[y] = (LTM_w_bias*val_set_bias[y] + LTM_w_prediction*LTM_prediction[y]) / (LTM_w_bias + LTM_w_prediction)
blah_LTM = sum_moo(LTM_weights)
for i in range(len(val_set)):
LTM_weights[i] = (LTM_weights[i]/blah_LTM)
LTM_output = [val_set_bias, LTM_prediction]
for g in range(len(val_set)):
aggregate_weights[g] = 0.75*STM_weights[g] + 0.25*LTM_weights[g]
blah_agg = sum_moo(aggregate_weights)
for i in range(len(val_set)):
aggregate_weights[i]= aggregate_weights[i]/blah_agg
output = weighted_choice(aggregate_weights)
return output
def adjust_weights(inp,out):
global memory_array_index, LTM_output, STM_output, streak, learning, STM_limit, LTM_limit
global STM_w_bias, STM_w_laystats, STM_w_prediction, LTM_w_bias,LTM_w_prediction
win = RPS(inp,out)
if win >= 0.50 and memory_array_index > streak and learning == 1:
for i in range(len(val_set)):
if val_set[i] == out:
if (LTM_output[0][i] > LTM_output[1][i]) and (LTM_w_prediction >= (0.05*win)):
LTM_w_bias += 0.05*win
LTM_w_prediction -= 0.05*win
if (LTM_output[0][i] < LTM_output[1][i]) and (LTM_w_bias >= (0.05*win)):
LTM_w_bias -= 0.05*win
LTM_w_prediction += 0.05*win
if ((STM_output[0][i] > STM_output[1][i]) and (STM_w_prediction >= 0.05*win)):
STM_w_bias += 0.10*win
STM_w_prediction -= 0.05*win
if ((STM_output[1][i] > STM_output[0][i]) and (STM_w_bias >= 0.05*win)):
STM_w_bias -= 0.05*win
STM_w_prediction -= 0.05*win
## if ((STM_output[0][i] > STM_output[1][i]) and (STM_w_laystats >= 0.05*win)) and ((STM_output[0][i] > STM_output[2][i]) and (STM_w_prediction >= 0.05*win)):
## STM_w_bias += 0.10*win
## STM_w_laystats -= 0.05*win
## STM_w_prediction -= 0.05*win
##
## if ((STM_output[1][i] > STM_output[0][i]) and (STM_w_bias >= 0.05*win) ) and ((STM_output[1][i] > STM_output[2][i]) and (STM_w_prediction >= 0.05*win)):
## STM_w_bias -= 0.05*win
## STM_w_laystats += 0.10*win
## STM_w_prediction -= 0.05*win
##
## if ((STM_output[2][i] > STM_output[0][i]) and (STM_w_bias >= 0.05*win)) and ((STM_output[2][i] > STM_output[1][i]) and (STM_w_laystats >= 0.05*win)):
## STM_w_bias -= 0.05*win
## STM_w_laystats -= 0.05*win
## STM_w_prediction += 0.10*win
return [LTM_w_bias,LTM_w_prediction, STM_w_bias, STM_w_laystats, STM_w_prediction]
else:
pass
if (input == 'R' or input == 'P' or input == 'S'):
winning.append(RPS(input,output))
a = random.randint(1,2)
if average(WLA_STM) > 0.50:
STM(input,output)
adjust_weights(input,output)
output = decision()
else:
STM(input,output)
adjust_weights(input,output)
decision()
output = random.choice(["R","P","S"])
else:
output = decision()