import scipy.io as sio
from CurrentThingsNeededtoRun import NA_Classifier
from CurrentThingsNeededtoRun.Transfer_Mat_From_Matlab import txmat
<<<<<<< HEAD
xtrain_BCDEFGHI = txmat('xtrain_BCDEFGHI_pow.mat','xtrain')
ytrain_BCDEFGHI = txmat('ytrain_BCDEFGHI_pow.mat','ytrain')
xtltrain_BCDEFGHI = txmat('xtltrain_BCDEFGHI_pow.mat','xtltrain')
xtest_BCDEFGHI = txmat('xtest_BCDEFGHI_pow.mat','xtest')
ytest_BCDEFGHI = txmat('ytest_BCDEFGHI_pow.mat','ytest')
xtltest_BCDEFGHI = txmat('xtltest_BCDEFGHI_pow.mat','xtltest')
xtrain_A = txmat('xtrain_A_pow.mat','xtrain') # we're gonna use this for testing'
ytrain_A = txmat('ytrain_A_pow.mat','ytrain')
xtltrain_A = txmat('xtltrain_A_pow.mat','xtltrain')
=======
xtrain_pow = sio.loadmat('xtrain_all_pow.mat')
xtrain_pow = xtrain_pow['xtrain']
>>>>>>> bd36724a187f99ba8e1e28e7167a1c6f578d48e7
print 'NA classifier pow training BCDEFGHI and A'
nu = [0.05, 0.1,.2,.3,.4, 0.5, 0.8]
for param in nu:
ystring,ystring1 = NA_Classifier.myclassify_NA(2, xtrain_BCDEFGHI, xtest_BCDEFGHI, xtltest_BCDEFGHI, xtrain_A, xtltrain_A, nuparam=param)
print 'for nu =' + str(param)
print 'results on BCDEFGHI testing set'
def lookatme():
xtesting = txmat('xtesting.mat','xtesting')
xtrunclength = txmat('xtrunclength.mat','xtrunclength')
xtrainallpow = txmat('xtrain_all_pow.mat','xtrain')
xtltrainallpow = txmat('xtltrain_all_pow.mat','xtltrain')
ytrainallpow = txmat('ytrain_all_pow.mat','ytrain')
# print "power results"
y,x,power9Class = FinalClassifier.myclassify_practice_set(1, xtrainallpow, ytrainallpow, xtltrainallpow, xtrunclength, xtesting)
# print y
# print power9Class
# print ""
# print FinalClassifier.predVec2Str(power9Class)
# print ""
xtrainallaud = txmat('xtrain_all_aud.mat','xtrain')
xtltrainallaud = txmat('xtltrain_all_aud.mat','xtltrain')
ytrainallaud = txmat('ytrain_all_aud.mat','ytrain')
# print 'audio results'
# print ""
y,x,audio9Class = FinalClassifier.myclassify_practice_set(1, xtrain = xtrainallaud, ytrain = ytrainallaud, xtltrain = xtltrainallaud, xtltest = xtrunclength, xtest = xtesting)
# print y
# print FinalClassifier.predVec2Str(audio9Class)
# print ""
ybintestallaud = txmat('ybintrain_all_aud.mat','ybintrain')
ybintestallpow = txmat('ybintrain_all_pow.mat','ybintrain')
y,audVpow = AudiovsPower.myclassify_AudPow(1, xtrainallaud, xtrainallpow, ybintestallaud, ybintestallpow, xtesting)
# print 'results from binary audio power classifier '
# print ""
# print y
# print ""
# print audVpow
grids = ['A','B','C','D','E','F','G','H','I']
pow1ClassMat = np.empty([len(grids),len(audVpow)])
#POWER ONE-CLASS
xtrain = []
xtltrain = []
for grid in grids:
xtrain.append(txmat('xtrain_' + grid + '_pow.mat','xtrain'))
xtltrain.append(txmat('xtltrain_' + grid + '_pow.mat','xtltrain'))
for i in range(len(xtrain)):
ystring,yvec = NA_Classifier.myclassify_oneclass(1, xtrain[i], xtesting, xtrunclength, nuparam = .05)
pow1ClassMat[i,:] = yvec
# print 'results on training set for training on ' + grids[i] + ' power '
# print ystring
# print pow1ClassMat
#AUDIO ONE-CLASS
# grids = ['A_18class1','B_18class1','C_18class1','D_18class1','E_18class1','F_18class1','G_18class1','H_18class1','I_18class1',
# 'A_18class2','B_18class2','C_18class2','D_18class2','E_18class2','F_18class2','G_18class2','H_18class2','I_18class2']
aud1ClassMat = np.empty([len(grids),len(audVpow)])
xtrain = []
xtltrain = []
for grid in grids:
xtrain.append(txmat('xtrain_' + grid + '_aud.mat','xtrain'))
xtltrain.append(txmat('xtltrain_' + grid + '_aud.mat','xtltrain'))
for i in range(len(xtrain)):
ystring,yvec = NA_Classifier.myclassify_oneclass(1, xtrain[i], xtesting, xtrunclength, nuparam = .1)
aud1ClassMat[i,:] = yvec
# print 'results on training set for training on ' + grids[i] + ' audio '
# print ystring
# print aud1ClassMat.shape
# we should run it for audio through 18 one class classifiers, one off each recording
xtrain_BD_aud = txmat('xtrain_BD_aud.mat','xtrain')
ytrain_BD_aud = txmat('ytrain_BD_aud.mat','ytrain')
xtltrain_BD_aud = txmat('xtltrain_BD_aud.mat','xtltrain')
y,x,BDsvm = FinalClassifier.myclassify_practice_set(1,xtrain_BD_aud,ytrain_BD_aud,xtltrain_BD_aud,xtrunclength,xtesting,grids='BD')
print BDsvm
xtrain_BE_aud = txmat('xtrain_BE_aud.mat','xtrain')
ytrain_BE_aud = txmat('ytrain_BE_aud.mat','ytrain')
xtltrain_BE_aud = txmat('xtltrain_BE_aud.mat','xtltrain')
y,x,BEsvm = FinalClassifier.myclassify_practice_set(1,xtrain_BE_aud,ytrain_BE_aud,xtltrain_BE_aud,xtrunclength,xtesting,grids='BE')
print BEsvm
#math time
finalclass = []
# print temp1classVec.shape
for i in range(len(audVpow)):
elem = audVpow[i]
if elem: #power
temp1classVec = pow1ClassMat[:,i]
if np.any(temp1classVec == 1):
finalclass.append(power9Class[i])
else:
finalclass.append(10) #N/A
else: #audio
temp1classVec = aud1ClassMat[:,i]
if np.any(temp1classVec == 1):
if audio9Class[i] == 2: #if we guess class B, use the BD Classifier to make sure it isn't D
if BDsvm[i] == 2: # BDsvm ==1 is B, 2 is D
finalclass.append(4) # if we guessed class D but the BD svm says it's B, append B
else:
finalclass.append(audio9Class[i])
elif audio9Class[i] == 5: #if we guess E, make sure it isn't B
if BEsvm[i] ==1:
finalclass.append(2)
else:
finalclass.append(audio9Class[i])
else:
finalclass.append(audio9Class[i])
# finalclass.append(audio9Class[i])
else:
finalclass.append(10) #N/A
# print finalclass
print FinalClassifier.predVec2Str(finalclass)