def sousProcessClassify(c,n_fold, loadingFolder):
#subprocess.call(["/media/lalil0u/New/software2/downloads/unpacked/svm-python-v204/svm_python_classify", "--m", "test", "-v", "3", "results/data_TEST_fold"+str(n_fold)+".pkl", "results/modelfile_"+c+"_"+str(n_fold)+".pkl"])
f = open(loadingFolder+"models/modelfile_"+c+"_"+str(n_fold)+".pkl", 'r')
mesPoids = pickle.load(f)
f.close()
#pdb.set_trace()
test_sol = classify.read_examples(loadingFolder+"data_TEST_fold"+str(n_fold)+".pkl")
total_loss = []; test_contenu = []
for (x,y) in test_sol:
test_contenu.append(classify.contenu(y))
ybar = classify.classify_example(x, mesPoids)
l = classify.loss(y, ybar)#attention ce n'est plus la hamming loss ms une liste de cinq chiffres
total_loss = l if total_loss == [] else np.vstack((total_loss, l))
return total_loss, test_contenu
def sousProcessClassify(sol, loadingFolder, loadingFile= None, i =None, n_f =None, n_big_f=None):
#subprocess.call(["/media/lalil0u/New/software2/downloads/unpacked/svm-python-v204/svm_python_classify", "--m", "test", "-v", "3", "results/data_TEST_fold"+str(n_fold)+".pkl", "results/modelfile_"+c+"_"+str(n_fold)+".pkl"])
if i==None:
f = open(os.path.join(loadingFolder,"modelfile_all.pkl"), 'r')
elif n_f==None:
c = "{:f}".format(10**i)
if n_big_f==None:
f = open(os.path.join(loadingFolder,"modelfile_all"+c+".pkl"), 'r')
else:
f = open(os.path.join(loadingFolder,"modelfile_all"+c+"_"+str(n_big_f)+".pkl"), 'r')
else:
c = "{:f}".format(10**i)
if n_big_f is not None:
f = open(os.path.join(loadingFolder,"modelfile_"+c+"_"+str(n_big_f)+'_'+str(n_f)+".pkl"), 'r')
else:
f = open(os.path.join(loadingFolder,"modelfile_"+c+"_"+str(n_f)+".pkl"), 'r')
mesPoids = pickle.load(f)
f.close()
new_sol = []
zz=0
if loadingFile == None:
for solu in sol.lstSolutions:
new_sol.append((solu, solu.truthVec()))
zz+=1
else:
new_sol = classify.read_examples(loadingFile)
for x,_ in new_sol:
r1=[]
try:
ybar = classify.classify_example(x, mesPoids)
except:
print "pbl de NaN"
x.truth = "PROBLEME DE NaN"
continue
else:
for k in range(len(ybar)):
r1.extend(ybar[k])
x.truth = r1
return new_sol
