def create_train_data():
prnt("matlab data로부터 noise를 첨가한 가상 실험 데이터를 생성합니다.")
inputs_fine = False
while not inputs_fine:
data_name = input("\n 데이터 이름 >>> ")
try :
ans_start = int(input("\n 실험변인 시작값(int) >>> "))
ans_end = int(input("\n 실험변인 종료값(int) >>> "))
ans_interval = int(input("\n 실험변인 간격(int) >>> "))
cond_mean = float(input("\n noise mean(float) >>> "))
cond_stdev = float(input("\n noise stdev(float) >>> "))
except :
prnt("올바른 숫자(int/float)를 입력해주세요.")
else :
inputs_fine = True
prnt("가상 실험 데이터를 생성합니다. 시간이 소요될 수 있습니다..")
trainData_Name = "trainData-"+dtm.datetime.now().strftime("%Y-%m%d-%H%M%S-%f")
tf = open("train_data/"+trainData_Name+".txt", "w")
tlf = open("train_data/trainData_List.txt", "a")
tlf.write(trainData_Name+"\n")
for param in range(ans_start, ans_end + ans_interval, ans_interval):
filepath = 'matlab_data/' + data_name +"_"+ str(param) + ".mat"
f = h5py.File(filepath, 'r')
dataArr = ""
tf.write(str(param) + "\n")
for label, raw_data in f.items():
dataArr = raw_data
for dataSet in raw_data:
tf.write(str(dataSet[0]) + "/" + str(dataSet[1] + np.random.normal(cond_mean, cond_stdev)) + "\n")
tf.write("\n")
prnt("가상 실험 데이터 생성이 완료되었습니다 : " + trainData_Name)
def estimation():
trlf = open("train_results/trainResults_List.txt", "r")
a = trlf.read().split('\n')
a.pop(len(a)-1)
a.sort(reverse=True)
prnt("다음 중 Esitimation을 진행할 train result set를 선택하세요.")
for idx, val in enumerate(a):
prnt(str(idx)+ " : " + val)
try :
dataSet_num = int(input("\n train result set 선택 >>> "))
except :
prnt("올바른 숫자를 입력하세요.")
tf = open("train_results/"+a[dataSet_num] + ".txt", "r")
b = tf.read().split('\n')
b.pop(len(b)-1)
b.sort(reverse=True)
prnt("다음 중 Esitimation을 진행할 학습 파라미터를 선택하세요.")
for idx, val in enumerate(b):
prnt(str(idx)+ " : " + val)
try :
param_num = int(input("\n 학습 파라미터 선택 >>> "))
except :
prnt("올바른 숫자를 입력하세요.")
param = b[param_num].split(' / ')
opt_slope = float(param[0].split(' ')[1])
opt_yint = float(param[1].split(' ')[1])
std = float(param[3].split(' ')[1])
e_p = float(param[4].split(' ')[1])
testf = open("test_data/testData_List.txt", "r")
c = testf.read().split('\n')
c.pop(len(c)-1)
prnt("다음 중 Esitimation의 대상이 되는 가상 test data를 선택하세요.")
for idx, val in enumerate(c):
prnt(str(idx)+ " : " + val)
try :
test_num = int(input("\n test data 선택 >>> "))
except :
prnt("올바른 숫자를 입력하세요.")
test_data_name = c[test_num]
tdf = open("test_data/"+ test_data_name +".txt", "r")
d = tdf.read().split('\n')
raw_data = d[1:]
raw_data.pop(len(raw_data)-1)
max_val = std * (1+e_p/100)
min_val = std * (1-e_p/100)
for data in raw_data:
temp_val = data.split("/")
if min_val<float(temp_val[1])<max_val:
time =float(temp_val[0])
break
estimated_val =(time - opt_yint) / opt_slope
prnt("추정 질량값 : " + str(estimated_val))
prnt("오차 : " + str( abs (100 * (float(d[0])-estimated_val)/float(d[0])) ) + " %" )
def train_regression():
tlf = open("train_data/trainData_List.txt", "r")
a = tlf.read().split('\n')
a.pop(len(a)-1)
a.sort(reverse=True)
prnt("다음 중 Regression을 진행할 실험 Data를 선택하세요.")
for idx, val in enumerate(a):
prnt(str(idx)+ " : " + val)
try :
dataSet_num = int(input("\n Train Data 선택 >>> "))
except :
prnt("올바른 숫자를 입력하세요.")
prnt("실험데이터를 읽어들입니다. 시간이 조금 소요됩니다..")
tf = open("train_data/"+a[dataSet_num] + ".txt", "r")
dt = []
dataArr = tf.read().split("\n\n")
dataArr.pop(len(dataArr)-1)
for data in dataArr:
data = data.split("\n")
dt.append(data)
prnt("실험데이터를 모두 읽어들였습니다.")
trainResults_Name = "trainResults-"+dtm.datetime.now().strftime("%Y-%m%d-%H%M%S-%f")
trf = open("train_results/"+trainResults_Name+".txt", "a")
trlf = open("train_results/trainResults_List.txt", "a")
trlf.write(trainResults_Name+"\n")
istesting = True
while istesting:
lam = 0
prnt("Regression을 위한 parameter들을 입력하세요.")
standard = float(input("\n 기준값 >>> "))
err_percent = float(input("\n 기준값과의 오차(%) >>> "))
regression_type = int(input("\n 1. Pure Linear Regression 2. L1 Normalization 3. L2 Normalization >>> "))
if (regression_type != 1):
lam = float(input("\n lambda for Normalization >>> "))
max_val = standard * (1+err_percent/100)
min_val = standard * (1-err_percent/100)
threshold_x = []
threshold_y = []
for data in dt:
x = data[0]
y = data[1:]
idx = 0
isFound = False
while idx < len(y) and (not isFound):
temp_val = y[idx].split("/")
if min_val<float(temp_val[1])<max_val:
isFound = True
threshold_x.append(int(x))
threshold_y.append( float(temp_val[0]))
idx += 1
y = np.array([threshold_y]).T
pi = np.array([ np.ones(len(threshold_x)).T,np.array(threshold_x).T]).T
if regression_type == 1:
optimal_parameters = inv(pi.T @ pi)@pi.T@y
elif regression_type == 2:
optimal_parameters = inv(pi.T @ pi)@(pi.T@y - (lam/2)* np.array([[1,1]]).T)
elif regression_type == 3:
optimal_parameters = inv(pi.T @ pi + lam* np.identity(2))@pi.T@y
else :
prnt("올바른 regression_type을 입력해주세요.")
optimal_values = optimal_parameters[1] * threshold_x + optimal_parameters[0]
mean = np.mean(y)
sstot = 0
ssres = 0
for y_idx, y_val in enumerate(y):
sstot += (y_val - mean)**2
ssres += (y_val - optimal_values[y_idx])**2
RSS = 1- (ssres/sstot)
trf.write("slope " + str(optimal_parameters[1][0]))
trf.write(" / y_intercept " + str(optimal_parameters[0][0]))
trf.write(" / RSS " + str(RSS) )
trf.write(" / standard "+ str(standard))
trf.write(" / err_percent " + str(err_percent))
trf.write(" / regression_type " + str(regression_type))
trf.write(" / lambda " + str(lam) + '\n')
plt.plot(threshold_x, threshold_y, 'go', threshold_x, optimal_values, 'r')
plt.title("Train : " + dtm.datetime.now().strftime("%Y.%m.%d/%H:%M:%S\n t = " +str(optimal_parameters[1]) +" m + "+str(optimal_parameters[0]) + ", RSS : " + str(RSS) ))
plt.ylabel("Reach Time")
plt.xlabel("Expermiental Mass")
plt.show()
finish = input("\n 종료하시려면 N, 다시하시려면 아무키나 누르세요 >>>")
if finish == "N":
istesting = False
def rm(name):
if os.path.exists(name):
cons.prnt('Removing %s ...'%name)
os.remove(name)
#!/usr/bin/python
import util as cons
try:
import distutils.ccompiler
cc = distutils.ccompiler.new_compiler()
cons.prnt('Compile...')
cc.compile(['codec.c'],include_dirs=['/usr/include/python3.1'],
extra_preargs=['-fPIC','-Wall','-shared','-Wstrict-prototypes']
)
cons.prnt('Link...')
cc.link(' ',['codec.o'],'codec.so',
libraries=['vorbis','speex']
)
cons.prnt('Okay')
except Exception as e:
cons.err('Compile Error: %s' %e)