def make_executable_Java ():
if not util.file_exists("solution.java"):
raise Exception("solution.java does not exist")
util.del_file("Main.java")
util.system("javac solution.java")
if not util.file_exists("Main.class"):
raise Exception("error in Java compilation")
def make_corrects_Java ():
for f in glob.glob("*.cor"):
util.del_file(f)
inps = sorted(glob.glob("*.inp"))
for inp in inps:
tst = os.path.splitext(inp)[0]
util.system("java Main <%s.inp >%s.cor" % (tst, tst))
def make_corrects_Python3 ():
for f in glob.glob("*.cor"):
util.del_file(f)
inps = sorted(glob.glob("*.inp"))
for inp in inps:
tst = os.path.splitext(inp)[0]
util.system("python3 solution.py <%s.inp >%s.cor" % (tst, tst))
def performance(origin_labels, predict_labels, deci_value, output=None, title=None, roc=False, bi_or_multi=0):
"""evaluations used to evaluate the performance of the model.
:param origin_labels: true values of the dataset.
:param predict_labels: predicted values of the dataset.
:param deci_value: decision values used for ROC and AUC.
:param output: the output file name of the ROC curve.
:param title: the title of the ROC curve.
:param roc: indicate whether to draw the ROC curve or not.
"""
if len(origin_labels) != len(predict_labels):
raise ValueError("The number of the original labels must equal to that of the predicted labels.")
if bi_or_multi == 0:
TP = 0.0
TN = 0.0
FP = 0.0
FN = 0.0
for i in range(len(origin_labels)):
if origin_labels[i] == 1.0 and predict_labels[i] == 1.0:
TP += 1.0
elif origin_labels[i] == 1.0 and predict_labels[i] == -1.0:
FN += 1.0
elif origin_labels[i] == -1.0 and predict_labels[i] == 1.0:
FP += 1.0
elif origin_labels[i] == -1.0 and predict_labels[i] == -1.0:
TN += 1.0
try:
SN = TP / (TP + FN)
except ZeroDivisionError:
SN = 0.0
try:
SP = TN / (FP + TN)
except ZeroDivisionError:
SP = 0.0
ACC = (TP + TN) / (TP + TN + FP + FN)
try:
MCC = (TP * TN - FP * FN) / math.sqrt((TP + FP) * (TP + FN) * (TN + FP) * (TN + FN))
except ZeroDivisionError:
MCC = 0.0
current_path = os.path.dirname(os.path.realpath(__file__))
roc_data_file = current_path + const.GEN_FILE_PATH + 'roc_data'
try:
AUC = plot_roc(deci_value, origin_labels, output, title, roc, roc_data_file)
except ZeroDivisionError:
AUC = 0.0
del_file(roc_data_file)
return ACC, MCC, AUC, SN, SP
elif bi_or_multi == 1:
correct_labels = 0.0
for elem in zip(origin_labels, predict_labels):
if elem[0] == elem[1]:
correct_labels += 1.0
ACC = correct_labels / len(origin_labels)
return ACC
def make_corrects_RunPython ():
for f in glob.glob("*.cor"):
util.del_file(f)
inps = sorted(glob.glob("*.inp"))
for inp in inps:
tst = os.path.splitext(inp)[0]
os.system("cat solution.py %s.inp > work.py" % tst)
util.system("python3 work.py >%s.cor" % (tst, ))
# additionally, create doctest-like session
if tst == 'sample':
python_doctest(tst)
def make_executable_C ():
handler = util.read_yml("handler.yml")
if handler["handler"] != "std":
raise Exception("unknown handler")
if not util.file_exists("solution.c"):
raise Exception("solution.c does not exist")
util.del_file("solution.exe")
util.system("%s %s solution.c -o solution.exe" % (cc, ccflags))
if not util.file_exists("solution.exe"):
raise Exception("error in C compilation")
def make_executable_GHC ():
handler = util.read_yml("handler.yml")
if handler["handler"] != "std":
raise Exception("unknown handler")
if not util.file_exists("solution.hs"):
raise Exception("solution.hs does not exist")
util.del_file("solution.exe")
util.system("ghc solution.hs -o solution.exe")
if not util.file_exists("solution.exe"):
raise Exception("error in GHC compilation")
def make_executable_PRO2 ():
util.del_file("solution.exe")
util.del_dir('compilation')
os.mkdir('compilation')
if util.file_exists("solution.cc"):
util.system('cp solution.cc compilation/program.cc')
elif util.file_exists("solution.hh"):
util.system('cp solution.hh compilation/program.hh')
else:
print "There is no solution.cc nor solution.hh"
util.system('cp public/* compilation')
util.system('cp private/* compilation')
os.chdir('compilation')
util.system("%s %s *.cc -o ../solution.exe" % (cxx, cxxflags))
os.chdir('..')
util.del_dir('compilation')
if not util.file_exists("solution.exe"):
raise Exception("solution.exe not created")
util.system("(cd public && tar cf ../public.tar *)")
util.system("(cd private && tar cf ../private.tar *)")
def make_corrects_RunHaskell ():
for f in glob.glob("*.cor"):
util.del_file(f)
inps = sorted(glob.glob("*.inp"))
for inp in inps:
tst = os.path.splitext(inp)[0]
util.copy_file("solution.hs", "work.hs")
if util.file_exists("judge.hs"):
os.system("cat judge.hs >> work.hs")
f = open("work.hs", "a")
print >>f, "main = do"
for line in open(tst+".inp").readlines():
line = line.rstrip()
if line.startswith("let "):
print >>f, " %s" % line
# elif line.startswith("deb "):
# print >>f, ' hPutStrLn stderr "%s"' % line
else:
print >>f, " print (%s)" % line
f.close()
util.system("runhaskell work.hs >%s.cor" % (tst, ))
def verify_program (program):
"""Verify that program compiles and gets AC for each test."""
# This implementation is not yet very functional, but works well in basic cases
# compile
program = os.path.splitext(program)[0]
if not util.file_exists(program+".cc"):
raise Exception("%s.cc does not exist" % program)
if not util.file_exists("handler.yml"):
raise Exception("handler.yml does not exist")
handler = util.read_yml("handler.yml")
if handler["handler"] != "std":
raise Exception("unknown handler")
util.del_file(program+".exe")
if util.file_exists("main.cc"):
raise Exception("not implemented yet")
if handler["source_modifier"] == "structs":
util.system("cat solution.cc main.cc > temporal.cc ; %s %s temporal.cc -o solution.exe ; rm temporal.cc" % (cxx, cxxflags))
else:
util.system("%s %s solution.cc main.cc -o solution.exe" % (cxx, cxxflags))
else:
if util.file_exists("solution.fallback"):
raise Exception("not implemented yet")
util.system("%s %s solution.cc -o solution.exe" % (cxx, cxxflags_fallback))
else:
util.system("%s %s %s.cc -o %s.exe" % (cxx, cxxflags, program, program))
if not util.file_exists(program+".exe"):
raise Exception(program+".exe not created")
# execute on tests
tests = sorted(glob.glob("*.inp"))
for test in tests:
test = os.path.splitext(test)[0]
os.system("./%s.exe < %s.inp > %s.out" % (program, test, test))
r = subprocess.call(["cmp", test+".out", test+".cor"])
if r: msg = "WA"
else: msg = "OK"
print "%s:\t\t%s" % (test, msg)
def make_executable_MakePRO2 ():
if not util.file_exists("solution"):
raise Exception("There is no solution directory")
if not util.file_exists("public"):
raise Exception("There is no public directory")
if not util.file_exists("private"):
raise Exception("There is no private directory")
util.del_file("solution.exe")
util.del_dir('compilation')
os.mkdir('compilation')
util.system('cp solution/* public/* private/* compilation')
os.chdir('compilation')
util.system("make")
util.system('cp program.exe ../solution.exe')
os.chdir('..')
util.del_dir('compilation')
if not util.file_exists("solution.exe"):
raise Exception("solution.exe not created")
util.system("(cd public && tar cf ../public.tar *)")
util.system("(cd private && tar cf ../private.tar *)")
util.system("(cd solution && tar cf ../solution.tar *)")
def make_executable_CPP ():
handler = util.read_yml("handler.yml")
if handler["handler"] != "std":
raise Exception("unknown handler")
if not util.file_exists("solution.cc"):
raise Exception("solution.cc does not exist")
util.del_file("solution.exe")
if util.file_exists("main.cc"):
if handler["source_modifier"] == "structs":
util.system("cat solution.cc main.cc > temporal.cc ; %s %s temporal.cc -o solution.exe ; rm temporal.cc" % (cxx, cxxflags))
else:
util.system("%s %s solution.cc main.cc -o solution.exe" % (cxx, cxxflags))
else:
if util.file_exists("solution.fallback"):
util.system("%s %s solution.cc -o solution.exe" % (cxx, cxxflags_fallback))
else:
util.system("%s %s solution.cc -o solution.exe" % (cxx, cxxflags))
if not util.file_exists("solution.exe"):
raise Exception("error in C++ compilation")
def make_corrects ():
"""Makes all correct files in the cwd."""
make_executable()
handler = util.read_yml("handler.yml")
if handler.get('compilers', '') == 'RunHaskell':
make_corrects_RunHaskell()
elif handler.get('compilers', '') == 'RunPython':
make_corrects_RunPython()
elif handler.get('solution', '') == 'Python3':
make_corrects_Python3()
elif handler.get('solution', '') == 'Java':
make_corrects_Java()
else:
if not util.file_exists("solution.exe"):
raise Exception("solution.exe does not exist")
for f in glob.glob("*.cor"):
util.del_file(f)
inps = sorted(glob.glob("*.inp"))
for inp in inps:
tst = os.path.splitext(inp)[0]
util.system("./solution.exe < %s.inp > %s.cor" % (tst, tst))
util.move_file(f, f2)
if util.file_exists('enunciat.tex'):
util.move_file('enunciat.tex', 'problem.es.tex')
if util.file_exists('*entrades*'):
util.move_file('entrades', 'inputs')
if util.file_exists('*entrada*'):
util.move_file('entrada', 'input')
if util.file_exists('puntets.txt'):
util.move_file('puntets.txt', 'scores.yml')
print " !!! AFEGEIX ELS sample-*/CANVIA ELS prova-* AL FITXER scores.yml !!!"
util.del_file('a.out')
for f in glob.glob('*.exe'):
util.del_file(f)
for f in glob.glob('*.ps'):
util.del_file(f)
for f in glob.glob('*.pdf'):
util.del_file(f)
util.move_file('problem.es.tex', 'problem.es.tex.latin1')
os.system('iconv -f latin1 -t utf8 < problem.es.tex.latin1 > problem.es.tex')
util.del_file('problem.es.tex.latin1')
handler = 'std'
def make_executable_Haskell ():
if not util.file_exists("solution.hs"):
raise Exception("solution.hs does not exist")
util.del_file("work")
util.del_file("work.hi")
util.del_file("work.o")
util.copy_file("solution.hs", "work.hs")
f = open("work.hs", "a")
print >>f, """main = do print "OK" """
f.close()
util.system("ghc -O3 work.hs")
if not util.file_exists("work"):
raise Exception("error in haskell compilation")
util.del_file("work")
util.del_file("work.hi")
util.del_file("work.o")
def cross_validation(label_list, vector_list, fold, svm_params, predict_params, bi_or_multi):
"""Do cross validation.
:param label_list: list of labels.
:param vector_list: list of vectors.
:param fold: the fold of cross validation.
"""
datasetsize = len(label_list)
result = dataset_split_cv(label_list, vector_list, fold)
if result == False:
return False
else:
split_vector_list, split_label_list = result
len_vector = len(split_vector_list)
len_label = len(split_label_list)
if len_vector != len_label:
print 'Error: The length of the labels is not equal to that of the vectors.'
return False
deci = []
original_labels = []
acc_list = []
mcc_list = []
auc_list = []
sn_list = []
sp_list = []
if bi_or_multi == 0:
if fold != datasetsize:
for i in range(len_vector):
train_vector_list = []
train_label_list = []
#test_vector_list = []
#test_label_list = []
test_vector_list = split_vector_list[i]
test_label_list = split_label_list[i]
original_labels.extend(test_label_list)
for j in range(len_vector):
if j != i:
train_vector_list.extend(split_vector_list[j])
train_label_list.extend(split_label_list[j])
m = svm_train(train_label_list, train_vector_list, svm_params)
p_label, p_acc, p_val = svm_predict(test_label_list, test_vector_list, m, predict_params)
labels = m.get_labels()
subdeci = [labels[0]*val[0] for val in p_val]
deci += subdeci
evals = performance(test_label_list, p_label, subdeci, bi_or_multi=bi_or_multi)
acc_list.append(evals[0])
mcc_list.append(evals[1])
auc_list.append(evals[2])
sn_list.append(evals[3])
sp_list.append(evals[4])
acc_average = sum(acc_list) / len(acc_list)
mcc_average = sum(mcc_list) / len(mcc_list)
auc_average = sum(auc_list) / len(auc_list)
sn_average = sum(sn_list) / len(sn_list)
sp_average = sum(sp_list) / len(sp_list)
label_all = []
for i in split_label_list:
label_all.extend(i)
check_gnuplot_exe()
roc_output = 'cross_validation.png'
title = 'cross validation'
current_path = os.path.dirname(os.path.realpath(__file__))
roc_data_file = current_path + const.GEN_FILE_PATH + 'roc_data'
plot_roc(deci, label_all, roc_output, title, True, roc_data_file)
del_file(roc_data_file)
dest_file = current_path + const.FINAL_RESULTS_PATH + roc_output
copy_file(roc_output, dest_file)
elif fold == datasetsize:
predicted_labels = []
#deci_list = []
for i in range(len_vector):
train_vector_list = []
train_label_list = []
#test_vector_list = []
#test_label_list = []
test_vector_list = split_vector_list[i]
test_label_list = split_label_list[i]
original_labels.extend(test_label_list)
for j in range(len_vector):
if j != i:
train_vector_list.extend(split_vector_list[j])
train_label_list.extend(split_label_list[j])
m = svm_train(train_label_list, train_vector_list, svm_params)
p_label, p_acc, p_val = svm_predict(test_label_list, test_vector_list, m, predict_params)
labels = m.get_labels()
subdeci = [labels[0]*val[0] for val in p_val]
deci += subdeci
predicted_labels.extend(p_label)
evals = performance(original_labels, predicted_labels, deci, bi_or_multi=bi_or_multi)
acc_average = evals[0]
mcc_average = evals[1]
auc_average = evals[2]
sn_average = evals[3]
sp_average = evals[4]
label_all = []
for i in split_label_list:
label_all.extend(i)
check_gnuplot_exe()
roc_output = 'cross_validation.png'
title = 'cross validation'
current_path = os.path.dirname(os.path.realpath(__file__))
roc_data_file = current_path + const.GEN_FILE_PATH + 'roc_data'
plot_roc(deci, label_all, roc_output, title, True, roc_data_file)
del_file(roc_data_file)
dest_file = current_path + const.FINAL_RESULTS_PATH + roc_output
copy_file(roc_output, dest_file)
#print acc_list
acc_re = 'ACC = %.4f' % acc_average
mcc_re = 'MCC = %.4f' % mcc_average
auc_re = 'AUC = %.4f' % auc_average
sn_re = 'Sn = %.4f' % sn_average
sp_re = 'Sp = %.4f\n' % sp_average
eval_re = [acc_re, mcc_re, auc_re, sn_re, sp_re]
print ('The cross validation results are as follows:')
print acc_re
print mcc_re
print auc_re
print sn_re
print sp_re
print "The ROC curve has been saved. You can check it here: "
if sys.platform.startswith('win'):
print dest_file.replace('/', '\\'), '\n'
else:
print dest_file.replace('\\', '/'), '\n'
result_file = current_path + const.FINAL_RESULTS_PATH + "cv_eval_results.txt"
with open(result_file, 'w') as f:
f.write('The cross validation results are as follows:\n')
for i in eval_re:
f.write(i)
f.write("\n")
prob_file = current_path + const.FINAL_RESULTS_PATH + "probability_values.txt"
with open(prob_file, 'w') as f:
for i, j in zip(original_labels, deci):
f.write(str(i))
f.write('\t')
f.write(str(j))
f.write("\n")
elif bi_or_multi == 1:
if fold != datasetsize:
for i in range(len_vector):
train_vector_list = []
train_label_list = []
#test_vector_list = []
#test_label_list = []
test_vector_list = split_vector_list[i]
test_label_list = split_label_list[i]
for j in range(len_vector):
if j != i:
train_vector_list.extend(split_vector_list[j])
train_label_list.extend(split_label_list[j])
m = svm_train(train_label_list, train_vector_list, svm_params)
p_label, p_acc, p_val = svm_predict(test_label_list, test_vector_list, m, predict_params)
labels = m.get_labels()
subdeci = [labels[0]*val[0] for val in p_val]
deci += subdeci
evals = performance(test_label_list, p_label, subdeci, bi_or_multi=bi_or_multi)
acc_list.append(evals)
acc_average = sum(acc_list) / len(acc_list)
elif fold == datasetsize:
predicted_labels = []
original_labels = []
for i in range(len_vector):
train_vector_list = []
train_label_list = []
#test_vector_list = []
#test_label_list = []
test_vector_list = split_vector_list[i]
test_label_list = split_label_list[i]
original_labels.extend(test_label_list)
for j in range(len_vector):
if j != i:
train_vector_list.extend(split_vector_list[j])
train_label_list.extend(split_label_list[j])
m = svm_train(train_label_list, train_vector_list, svm_params)
p_label, p_acc, p_val = svm_predict(test_label_list, test_vector_list, m, predict_params)
labels = m.get_labels()
subdeci = [labels[0]*val[0] for val in p_val]
deci += subdeci
predicted_labels.extend(p_label)
evals = performance(original_labels, predicted_labels, deci, bi_or_multi=bi_or_multi)
acc_average = evals
print ('The cross validation results are as follows:')
print 'ACC = %.4f' % acc_average
#def cv_jackknife(label_list, vector_list, fold, svm_params, predict_params, bi_or_multi):
"""Do jackknife cross validation.
print("parse " + k)
# 将指定Excel文件转换成json文本
def excel2json(xlsx_path):
# load excel data
xlsx_data = open_excel(xlsx_path)
# 获取所有sheet
nsheets = xlsx_data.nsheets
for idx in range(nsheets):
sheet = xlsx_data.sheets()[idx]
if sheet.nrows > 0 or sheet.ncols > 0:
sheet2json(sheet)
# 删除原文件
util.del_file(LOC_CFG_PATH)
files = os.listdir(XLSX_PATH)
for f in files:
name, ext = os.path.splitext(f)
if (ext == '.xls' or ext == '.xlsx'):
if name.find("~") >= 0:
print("ignore file ", f)
continue
print("export " + f)
excel2json(XLSX_PATH + "/" + f)
print("localization Done")
def cross_validation(label_list, vector_list, fold, svm_params, predict_params,
bi_or_multi):
"""Do cross validation.
:param label_list: list of labels.
:param vector_list: list of vectors.
:param fold: the fold of cross validation.
"""
result = dataset_split_cv(label_list, vector_list, fold)
if result == False:
return False
else:
split_vector_list, split_label_list = result
len_vector = len(split_vector_list)
len_label = len(split_label_list)
if len_vector != len_label:
print 'Error: The length of the labels is not equal to that of the vectors.'
return False
deci = []
acc_list = []
mcc_list = []
auc_list = []
sn_list = []
sp_list = []
if bi_or_multi == 0:
for i in range(len_vector):
train_vector_list = []
train_label_list = []
#test_vector_list = []
#test_label_list = []
test_vector_list = split_vector_list[i]
test_label_list = split_label_list[i]
for j in range(len_vector):
if j != i:
train_vector_list.extend(split_vector_list[j])
train_label_list.extend(split_label_list[j])
m = svm_train(train_label_list, train_vector_list, svm_params)
p_label, p_acc, p_val = svm_predict(test_label_list,
test_vector_list, m,
predict_params)
labels = m.get_labels()
subdeci = [labels[0] * val[0] for val in p_val]
deci += subdeci
evals = performance(test_label_list,
p_label,
subdeci,
bi_or_multi=bi_or_multi)
acc_list.append(evals[0])
mcc_list.append(evals[1])
auc_list.append(evals[2])
sn_list.append(evals[3])
sp_list.append(evals[4])
acc_average = sum(acc_list) / len(acc_list)
mcc_average = sum(mcc_list) / len(mcc_list)
auc_average = sum(auc_list) / len(auc_list)
sn_average = sum(sn_list) / len(sn_list)
sp_average = sum(sp_list) / len(sp_list)
label_all = []
for i in split_label_list:
label_all.extend(i)
check_gnuplot_exe()
roc_output = 'cross_validation.png'
title = 'cross validation'
current_path = os.path.dirname(os.path.realpath(__file__))
roc_data_file = current_path + const.GEN_FILE_PATH + 'roc_data'
plot_roc(deci, label_all, roc_output, title, True, roc_data_file)
del_file(roc_data_file)
dest_file = current_path + const.FINAL_RESULTS_PATH + roc_output
copy_file(roc_output, dest_file)
print('The cross validation results are as follows:')
print 'ACC = %.4f' % acc_average
print 'MCC = %.4f' % mcc_average
print 'AUC = %.4f' % auc_average
print 'Sn = %.4f' % sn_average
print 'Sp = %.4f\n' % sp_average
print "The ROC curve has been saved. You can check it here: "
if sys.platform.startswith('win'):
print dest_file.replace('/', '\\'), '\n'
else:
print dest_file.replace('\\', '/'), '\n'
elif bi_or_multi == 1:
for i in range(len_vector):
train_vector_list = []
train_label_list = []
#test_vector_list = []
#test_label_list = []
test_vector_list = split_vector_list[i]
test_label_list = split_label_list[i]
for j in range(len_vector):
if j != i:
train_vector_list.extend(split_vector_list[j])
train_label_list.extend(split_label_list[j])
m = svm_train(train_label_list, train_vector_list, svm_params)
p_label, p_acc, p_val = svm_predict(test_label_list,
test_vector_list, m,
predict_params)
labels = m.get_labels()
subdeci = [labels[0] * val[0] for val in p_val]
deci += subdeci
evals = performance(test_label_list,
p_label,
subdeci,
bi_or_multi=bi_or_multi)
acc_list.append(evals)
acc_average = sum(acc_list) / len(acc_list)
print('The cross validation results are as follows:')
print 'ACC = %.4f' % acc_average