def naive_bayes_with_lda():
train, train_target, test, test_target = load_polluted_spambase()
print "Train data: %s, Train Label: %s" % (train.shape, train_target.shape)
print "Test data: %s, Test Label: %s" % (test.shape, test_target.shape)
start = timeit.default_timer()
lda = LDA(n_components=100)
train = lda.fit_transform(train, train_target)
test = lda.transform(test)
print lda
print "Train data: %s, Train Label: %s" % (train.shape, train_target.shape)
print "Test data: %s, Test Label: %s" % (test.shape, test_target.shape)
cf = GaussianNaiveBayes()
cf.fit(train, train_target)
raw_predicts = cf.predict(test)
predict_class = cf.predict_class(raw_predicts)
cm = confusion_matrix(test_target, predict_class)
print "confusion matrix: TN: %s, FP: %s, FN: %s, TP: %s" % (
cm[0, 0], cm[0, 1], cm[1, 0], cm[1, 1])
er, acc, fpr, tpr = confusion_matrix_analysis(cm)
print 'Error rate: %f, accuracy: %f, FPR: %f, TPR: %f' % (er, acc, fpr,
tpr)
stop = timeit.default_timer()
print "Total Run Time: %s secs" % (stop - start)
def naive_bayes_with_lda():
train, train_target, test, test_target = load_polluted_spambase()
print "Train data: %s, Train Label: %s" % (train.shape, train_target.shape)
print "Test data: %s, Test Label: %s" % (test.shape, test_target.shape)
start = timeit.default_timer()
lda = LDA(n_components=100)
train = lda.fit_transform(train, train_target)
test = lda.transform(test)
print lda
print "Train data: %s, Train Label: %s" % (train.shape, train_target.shape)
print "Test data: %s, Test Label: %s" % (test.shape, test_target.shape)
cf = GaussianNaiveBayes()
cf.fit(train, train_target)
raw_predicts = cf.predict(test)
predict_class = cf.predict_class(raw_predicts)
cm = confusion_matrix(test_target, predict_class)
print "confusion matrix: TN: %s, FP: %s, FN: %s, TP: %s" % (cm[0, 0], cm[0, 1], cm[1, 0], cm[1, 1])
er, acc, fpr, tpr = confusion_matrix_analysis(cm)
print "Error rate: %f, accuracy: %f, FPR: %f, TPR: %f" % (er, acc, fpr, tpr)
stop = timeit.default_timer()
print "Total Run Time: %s secs" % (stop - start)
def runLDA(all_kmer_vectors_array,labels):
sklearn_lda = LDA(n_components=4)
X = np.array(all_kmer_vectors_array)
y = np.array(labels)
X_lda_sklearn = sklearn_lda.fit_transform(X,y)
print(X_lda_sklearn)
return X_lda_sklearn
def lda_scikit():
df_wine = pd.read_csv('https://archive.ics.uci.edu/ml/machine-learning-databases/wine/wine.data', header=None)
df_wine.columns = ['Class label', 'Alcohol', 'Malic acid', 'Ash',
'Alcalinity of ash', 'Magnesium', 'Total phenols',
'Flavanoids', 'Nonflavanoid phenols', 'Proanthocyanins',
'Color intensity', 'Hue', 'OD280/OD315 of diluted wines', 'Proline']
X, y = df_wine.iloc[:, 1:].values, df_wine.iloc[:, 0].values
X_train, X_test, y_train, y_test = \
train_test_split(X, y, test_size=0.3, random_state=0)
sc = StandardScaler()
X_train_std = sc.fit_transform(X_train)
X_test_std = sc.transform(X_test)
pdb.set_trace()
lda = LDA(n_components=3)
X_train_lda = lda.fit_transform(X_train_std, y_train)
lr = LogisticRegression()
lr = lr.fit(X_train_lda, y_train)
plot_decision_regions(X_train_lda, y_train, classifier=lr)
plt.xlabel('LD 1')
plt.ylabel('LD 2')
plt.legend(loc='lower left')
plt.tight_layout()
plt.savefig(PL5 + 'lda_scikit.png', dpi=300)
plt.close()
X_test_lda = lda.transform(X_test_std)
plot_decision_regions(X_test_lda, y_test, classifier=lr)
plt.xlabel('LD 1')
plt.ylabel('LD 2')
plt.legend(loc='lower left')
plt.tight_layout()
plt.savefig(PL5 + 'lda_scikit_test.png', dpi=300)
def combine_lda_pca(X, y):
sklearn_lda = LDA(n_components=2)
X_lda_sklearn = sklearn_lda.fit_transform(X, y)
sklearn_pca = sklearnPCA(n_components=2) #PCA
X_ldapca_sklearn = sklearn_pca.fit_transform(X_lda_sklearn)
plot_scikit_lda(X_ldapca_sklearn,
title='LDA+PCA via scikit-learn',
mirror=(-1))
def _lda(self):
""" performs linear discriminant analysis of the data"""
if not self.target:
print "target has not been set, it is required for LDA"
return
lda = LDA(n_components=self.n)
self.transformed_data = lda.fit_transform(
self.data, self.target).transform(self.data)
def reduceDimensionLDA(mat, k): print mat.shape labels = mat[:, -1] mat = mat[:, :-1] lda = LDA(n_components = k) data = lda.fit_transform(mat, labels) data = addLabels(data, labels) print data return data
def lda(df, samples, sample_labels, plot_name='lda_plot.png'):
df = df.copy()
df = df.transpose()
df = df.ix[samples]
df_nrm = normalize_min_max(df)
X = df_nrm.values
label_dict, y = encode_labels(sample_labels)
ldas = LDA(n_components=2)
X_lda = ldas.fit_transform(X, y)
plot_scikit_lda(X_lda, y, label_dict, samples)
def lda_decompose(dataset, n):
lda = LDA(n_components=n)
reduced_features = lda.fit_transform(dataset.all.features,
dataset.all.target)
training_size = dataset.training_size
training = Data(reduced_features[:training_size, :],
dataset.all.target[:training_size])
testing = Data(reduced_features[training_size:, :],
dataset.all.target[training_size:])
return DataSet(training, testing)
def lda(df, samples, sample_labels, plot_name='lda_plot.png'):
df = df.copy()
df = df.transpose()
df = df.ix[samples]
df_nrm = normalize_min_max(df)
X = df_nrm.values
label_dict, y = encode_labels(sample_labels)
ldas = LDA(n_components=2)
X_lda = ldas.fit_transform(X, y)
plot_scikit_lda(X_lda, y, label_dict, samples)
def execute(self,i,j):
# dim_red = LDA()
# dim_red.fit_transform(self.x_train, self.y_train)
# with open('dumped_dim_red_'+str(i)+'.pkl', 'wb') as fid:
# cPickle.dump(dim_red, fid)
# x_train = dim_red.transform(self.x_train)
# x_test = dim_red.transform(self.y_train)
# stat_obj = self.stat_class() # reflection bitches
# stat_obj.train(x_train, x_test)
# print len(x_train)
# with open('dumped_'+str(j)+'_'+str(i)+'.pkl', 'wb') as fid:
# cPickle.dump(stat_obj, fid)
kf = KFold(len(self.x_train), n_folds=self.k_cross)
own_kappa = []
for train_idx, test_idx in kf:
# print train_idx,test_idx
# exit(0)
x_train, x_test = self.x_train[train_idx], self.x_train[test_idx]
y_train, y_test = self.y_train[train_idx], self.y_train[test_idx]
dim_red = LDA()
x_train = dim_red.fit_transform(x_train, y_train)
# with open('dumped_dim_red_'+str(i)+'.pkl', 'wb') as fid:
# cPickle.dump(dim_red, fid)
# with open('dumped_dim_red_'+str(i)+'.pkl', 'rb') as fid:
# dim_red=cPickle.load(fid)
x_test = dim_red.transform(x_test)
# with open('dumped_'+str(j)+'_'+str(i)+'.pkl', 'rb') as fid:
# stat_obj=cPickle.load(fid)
# x_train = dim_red.transform(x_train)
# x_test = dim_red.transform(x_test)
stat_obj = self.stat_class() # reflection bitches
stat_obj.train(x_train,y_train)
# with open('dumped_'+str(j)+'_'+str(i)+'.pkl', 'wb') as fid:
# cPickle.dump(stat_obj, fid)
# with open('dumped_'+str(j)+'_'+str(i)+'.pkl', 'rb') as fid:
# stat_obj=cPickle.load(fid)
y_pred = [ 0 for i in xrange(len(y_test)) ]
for i in range(len(x_test)):
# print len(x_test[i])
val = int(np.round(stat_obj.predict(x_test[i])))
if val > self.range_max: val = self.range_max
if val < self.range_min: val = self.range_min
y_pred[i] = [val]
y_pred = np.matrix(y_pred)
cohen_kappa_rating = own_wp.quadratic_weighted_kappa(y_test,y_pred,self.range_min,self.range_max)
self.values.append(cohen_kappa_rating)
return str(sum(self.values)/self.k_cross)
def get_LDA_performance(test_df, X_std, y):
X_test = test_df.ix[:, 'x.1':'x.10'].values
X_std_test = StandardScaler().fit_transform(X_test)
y_test = test_df.ix[:, 'y'].values
lda_scores_training = []
lda_scores_test = []
qda_scores_training = []
qda_scores_test = []
knn_scores_training = []
knn_scores_test = []
for d in range(1, 11):
lda = LDA(n_components=d)
Xred_lda_training = lda.fit_transform(X_std, y)
Xred_lda_test = lda.transform(X_std_test)
lda_model = LDA()
lda_model.fit(Xred_lda_training, y)
qda_model = QDA()
qda_model.fit(Xred_lda_training, y)
knn_model = KNeighborsClassifier(n_neighbors=10)
knn_model.fit(Xred_lda_training, y)
lda_scores_training.append(1 - lda_model.score(Xred_lda_training, y))
lda_scores_test.append(1 - lda_model.score(Xred_lda_test, y_test))
qda_scores_training.append(1 - qda_model.score(Xred_lda_training, y))
qda_scores_test.append(1 - qda_model.score(Xred_lda_test, y_test))
knn_scores_training.append(1 - knn_model.score(Xred_lda_training, y))
knn_scores_test.append(1 - knn_model.score(Xred_lda_test, y_test))
plt.plot(range(10), lda_scores_training, 'r--', label="Train data")
plt.plot(range(10), lda_scores_test, 'b--', label="Test data")
plt.title("LDA vs LDA")
plt.xlabel('k')
plt.ylabel('Score')
plt.show()
plt.plot(range(10), qda_scores_training, 'r--', label="Train data")
plt.plot(range(10), qda_scores_test, 'b--', label="Test data")
plt.title("QDA vs LDA")
plt.show()
plt.plot(range(10), knn_scores_training, 'r--', label="Train data")
plt.plot(range(10), knn_scores_test, 'b--', label="Test data")
plt.title("KNN vs LDA")
plt.show()
class LDA(AbstractProjection):
def __init__(self, **kw):
super(LDA, self).__init__()
self.lda = ScikitLDA(**kw)
def train(self, features, labels):
red_feats = self.lda.fit_transform(features, labels)
self.V = np.std(red_feats, axis=0)
def project(self, feats, whiten=True):
lda_feats = self.lda.transform(feats)
if whiten:
lda_feats /= self.V
return lda_feats
def execute(self,i,j):
x_train= self.x_train
y_train= self.y_train
dim_red = LDA()
x_train = dim_red.fit_transform(x_train, y_train)
with open('dumped_dim_red_'+str(i)+'.pkl', 'wb') as fid:
cPickle.dump(dim_red, fid)
stat_obj = self.stat_class() # reflection bitches
stat_obj.train(x_train,y_train)
with open('dumped_'+str(j)+'_'+str(i)+'.pkl', 'wb') as fid:
cPickle.dump(stat_obj, fid)
kf = KFold(len(self.x_train), n_folds=self.k_cross)
own_kappa = []
for train_idx, test_idx in kf:
# print train_idx,test_idx
# exit(0)
x_train, x_test = self.x_train[train_idx], self.x_train[test_idx]
y_train, y_test = self.y_train[train_idx], self.y_train[test_idx]
dim_red = LDA()
x_train = dim_red.fit_transform(x_train, y_train)
x_test = dim_red.transform(x_test)
stat_obj = self.stat_class() # reflection bitches
stat_obj.train(x_train,y_train)
y_pred = [ 0 for i in xrange(len(y_test)) ]
for i in range(len(x_test)):
val = int(np.round(stat_obj.predict(x_test[i])))
if val > self.range_max: val = self.range_max
if val < self.range_min: val = self.range_min
y_pred[i] = [val]
y_pred = np.matrix(y_pred)
cohen_kappa_rating = own_wp.quadratic_weighted_kappa(y_test,y_pred,self.range_min,self.range_max)
self.values.append(cohen_kappa_rating)
return sum(self.values)/self.k_cross
def lda_data(X, y, n_components=2, num_data_points=-1):
lda = LDA(n_components=n_components)
if num_data_points > 0:
X = X[:num_data_points, :]
y = y[:num_data_points]
print "Performing mapping"
start = timeit.default_timer()
mapped = lda.fit_transform(X, y)
end = timeit.default_timer()
print "Mapping completed in %f seconds" % (end - start)
return mapped, lda
def get_LDA(X_std, y):
sklearn_lda = LDA(n_components=2)
Xred_lda = sklearn_lda.fit_transform(X_std, y)
cmap = plt.cm.get_cmap('Accent')
mclasses = (1, 2, 3, 4, 5, 6, 7, 8, 9)
mcolors = [cmap(i) for i in np.linspace(0, 1, 10)]
plt.figure(figsize=(12, 8))
for lab, col in zip(mclasses, mcolors):
plt.scatter(Xred_lda[y == lab, 0],
Xred_lda[y == lab, 1],
label=lab,
c=col)
plt.xlabel('LDA/Fisher Direction 1')
plt.ylabel('LDA/Fisher Direction 2')
leg = plt.legend(loc='upper right', fancybox=True)
plt.show()
def with_lda(X_train_std, y_train, X_test_std, y_test):
from sklearn.lda import LDA
lda = LDA(n_components=2)
X_train_lda = lda.fit_transform(X_train_std, y_train)
lr = LogisticRegression()
lr = lr.fit(X_train_lda, y_train)
plot_decision_regions(X_train_lda, y_train, classifier=lr)
plot.xlabel('LD 1')
plot.ylabel('LD 2')
plt.legend(loc='lower left')
plt.show()
X_test_lda = lda.transform(X_test_std)
plot_decision_regions(X_test_lda, y_test, classifier=lr)
plot.xlabel('LD 1')
plot.ylabel('LD 2')
plt.legend(loc='lower left')
plt.show()
def execute(self):
kf = KFold(len(self.x_train), n_folds=self.k_cross)
own_kappa = []
for train_idx, test_idx in kf:
x_train, x_test = self.x_train[train_idx], self.x_train[test_idx]
y_train, y_test = self.y_train[train_idx], self.y_train[test_idx]
dim_red = LDA()
x_train = dim_red.fit_transform(x_train, y_train)
x_test = dim_red.transform(x_test)
stat_obj = self.stat_class() # reflection bitches
stat_obj.train(x_train,y_train)
y_pred = [ 0 for i in xrange(len(y_test)) ]
for i in range(len(x_test)):
val = int(np.round(stat_obj.predict(x_test[i])))
if val > self.range_max: val = self.range_max
if val < self.range_min: val = self.range_min
y_pred[i] = [val]
y_pred = np.matrix(y_pred)
cohen_kappa_rating = own_wp.quadratic_weighted_kappa(y_test,y_pred,self.range_min,self.range_max)
self.values.append(cohen_kappa_rating)
return str(sum(self.values)/self.k_cross)
from Wine import getWineData
from Util import plot_decision_regions
from sklearn.lda import LDA
from sklearn.linear_model import LogisticRegression
import numpy as np
import matplotlib.pyplot as plt
X_train_std, X_test_std, y_train, y_test = getWineData()
lda = LDA(n_components=2)
X_train_lda = lda.fit_transform(X_train_std, y_train)
lr = LogisticRegression()
lr.fit(X_train_lda, y_train)
plot_decision_regions(X_train_lda, y_train, classifier=lr)
plt.xlabel('LD 1')
plt.ylabel('LD 2')
plt.legend(loc='lower left')
plt.show()
# newdata = normdata # for i in range(5): # print newdata[i] print "data done" print "logistic initialized" # clf.fit(data[:,:-1], data[:,-1]) print "fitted data" skf = StratifiedKFold(data[:,-1], n_folds=10, shuffle=True) output =[] finalscore = 0 counter = 0 for train, test in skf: counter = counter + 1 newdata = prj.fit_transform([ normdata[i][:] for i in train ],[ data[i][-1] for i in train ]) newtestdata = prj.transform([ normdata[i][:] for i in test ]) clf = GradientBoostingClassifier(warm_start = True) clf = clf.fit(newdata, [ data[i][-1] for i in train ]) prediction = clf.predict(newtestdata) # pred = [] # for i in prediction: # if(i > 1.5): # pred.append(2) # else: # pred.append(1) finalscore = finalscore + score.get_score( prediction , [ data[i][-1] for i in test ]) print "done" # score = cross_val_score(clf, newdata[:,:], data[:,-1], cv = 5, scoring = 'get_score') # print "in scores" # for i in score:
except:
pass
for i in range(len(s2.split('\n'))):
Test.append([])
for j in s2.split('\n')[i].split(','):
try:
Test[i].append(float(j))
except:
pass
pca=PCA(n_components=2)
x=pca.fit_transform(numpy.array(Data[0:len(Data)-1]))
y=pca.fit_transform(numpy.array(Test[0:len(Test)-1]))
regr=linear_model.LinearRegression()
regr.fit(x,1000*[0]+1000*[1])
clf=LDA(n_components=2)
z=clf.fit_transform(numpy.array(Data[0:len(Data)-1]),1000*[0]+1000*[1])
New=clf.predict(Test[0:len(Test)-1])
A=numpy.transpose(x)
B=numpy.dot(A,x)
C=numpy.dot(inv(B),A)
D=numpy.dot(C,1000*[-1]+1000*[1])
E=numpy.transpose(D)
F=numpy.dot(y,D)
outx=[]
for i in F:
if i>=0:
outx.append(1)
else:
outx.append(0)
outp=[]
for i in regr.predict(y):
def combine_lda_pca(X, y):
sklearn_lda = LDA(n_components=2)
X_lda_sklearn = sklearn_lda.fit_transform(X, y)
sklearn_pca = sklearnPCA(n_components=2) #PCA
X_ldapca_sklearn = sklearn_pca.fit_transform(X_lda_sklearn)
plot_scikit_lda(X_ldapca_sklearn, title='LDA+PCA via scikit-learn', mirror=(-1))
import pandas as pd from sklearn import datasets from sklearn.decomposition import PCA from sklearn.lda import LDA # Load data iris = datasets.load_iris() idata = iris.data itarget = iris.target species = iris.target_names iris_df = pd.DataFrame(idata, columns = ['Sepal Length', 'Sepal Width', 'Petal Length', 'Petal Width']) iris_df['Species'] = itarget # PCA pca = PCA(n_components = 2) reduced_idata = pca.fit_transform(idata) # LDA lda = LDA(n_components = 2) reduced_itarget = lda.fit_transform(idata, itarget)
from sklearn.datasets import load_iris
import numpy as np
iris = load_iris()
print 'iris.data[-10:]\n', iris.data[-10:]
print 'iris.target[-10:]\n', iris.target[-10:]
print 'iris.data.shape:', iris.data.shape
from sklearn.lda import LDA
# Data dimension reduction
lda = LDA() # Default n_components setting, max C-1
lda_result1 = lda.fit_transform(iris.data, iris.target)
print 'LDA result 1:', lda_result1.shape
lda = LDA(n_components=1)
lda_result2 = lda.fit_transform(iris.data, iris.target)
print 'LDA result 2:', lda_result2.shape
# Visualization
import matplotlib.pyplot as plt
plt.subplot(1,2,1)
plt.scatter(lda_result1[iris.target==0, 0], lda_result1[iris.target==0, 1], color='r')
plt.scatter(lda_result1[iris.target==1, 0], lda_result1[iris.target==1, 1], color='g')
plt.scatter(lda_result1[iris.target==2, 0], lda_result1[iris.target==2, 1], color='b')
plt.title('LDA on iris (1)')
plt.subplot(1,2,2)
plt.stem(lda_result2)
plt.title('LDA on iris (2)')
plt.show()
# A cada dato se le asigna la clase y un color para diferenciarlas entre si
for lab, col in zip(mclasses,mcolors):
plt.scatter(Xred_pca[y==lab, 0],Xred_pca[y==lab, 1],label=lab,c=col)
# Se configuran las etiquetas
plt.xlabel('Principal Component 1')
plt.ylabel('Principal Component 2')
leg = plt.legend(loc='upper right', fancybox=True)
######## Pregunta (d) ############################################################
# Se utiliza LDA con dos dimensiones
sklearn_lda = LDA(n_components=2)
# Se ajusta a los datos de entrenamiento
Xred_lda = sklearn_lda.fit_transform(X_std,y)
# Se escoge la paleta de colores
cmap = plt.cm.get_cmap('hsv')
# Se definen las clases
mclasses=(1,2,3,4,5,6,7,8,9)
mcolors = [cmap(i) for i in np.linspace(0,1,10)]
# Se establece el tamanno de la figura
plt.figure(figsize=(12, 8))
# A cada dato se le asigna la clase y un color para diferenciarlas entre si
for lab, col in zip(mclasses,mcolors):
plt.scatter(Xred_lda[y==lab, 0],Xred_lda[y==lab, 1],label=lab,c=col)
with closing(open(path.join(__file__, '..', 'data', 'X_train.pkl'), 'rb')) as pkl:
data = cPickle.load(pkl)
with closing(open(path.join(__file__, '..', 'data', 'y_train.pkl'), 'rb')) as pkl:
target = cPickle.load(pkl)
vectorizer = CountVectorizer(max_df=.75, min_df = 2, ngram_range=(1,2))
X=vectorizer.fit_transform(data, array(target))
transformer = TfidfTransformer()
X = transformer.fit_transform(X, array(target))
#hv = HashingVectorizer()
clf = LDA()
clf.fit_transform(X.toarray(), array(target), store_covariance=True)
for name,obj in [('saclings.txt', clf.scalings_), ('coef.txt', clf.coef_),
('covariance.txt', clf.covariance_),('xbar.txt', clf.xbar_),
('means.txt', clf.means_)]:
with closing(open(path.join(opt['out_path'], name), 'wb')) as out:
print 'saving %s' % name
for row in obj:
out.write(str(row)+'\r\n')
print 'priors'
print clf.priors
del X
del data
del target
#kernel PCA keeping 300 components
kpca = KernelPCA(kernel="rbf",n_components=300 , gamma=1)
X_kpca = kpca.fit_transform(X_train)
X_test = kpca.transform(X_test)
print (kpca)
print(X_kpca.shape)
#lda for dimensionality reduction. It should keep [classes-1] components.
lda = LDA()
print (lda)
X_lda = lda.fit_transform(X_kpca,y_train)
X_test = lda.transform(X_test)
print(X_lda.shape)
#kNN classification
start = int(round(time.time() * 1000))
clf = neighbors.KNeighborsClassifier(n_neighbors=5)
clf.fit(X_lda, y_train)
print (clf)
print("---------(5) Cross validation accuracy--------")
print(cross_validation.cross_val_score(clf, X_lda,y_train, cv=5))
acc, clusters = run_clustering(X_new)
print "average EM score after X modified with ICA", n, "components, clusters =", clusters, "silhouette score =", acc
if dralg == 'rp':
#######################################################
######## KMeans after Sparse Random Projection ########
#######################################################
for n in range(1, len(df.columns) + 1):
# create the random projection
sp = SparseRandomProjection(n_components = n)
X_new = sp.fit_transform(X)
acc, clusters = run_clustering(X_new)
print "average EM score after X modified with Random Projectsion", n, "components, clusters =", clusters, "silhouette score =", acc
if dralg == 'lda':
##################################
######## KMeans after LDA ########
##################################
for n in range(1, len(df.columns) + 1):
for solver in ['svd', 'eigen']:
# create the random projection
lda = LDA(n_components = n, solver = solver)
X_new = lda.fit_transform(X, y)
acc, clusters = run_clustering(X_new)
print "average EM score after X modified with LDA", n, "components, clusters =", clusters, "silhouette score =", acc
plt.show()
def execute(self, i, j):
global save1
global save2
jk = i
# print type(jk)
# dim_red = LDA()
# dim_red.fit_transform(self.x_train, self.y_train)
# with open('dumped_dim_red_'+str(i)+'.pkl', 'wb') as fid:
# cPickle.dump(dim_red, fid)
# x_train = dim_red.transform(self.x_train)
# x_test = dim_red.transform(self.y_train)
# stat_obj = self.stat_class() # reflection bitches
# stat_obj.train(x_train, x_test)
# print len(x_train)
# with open('dumped_'+str(j)+'_'+str(i)+'.pkl', 'wb') as fid:
# cPickle.dump(stat_obj, fid)
# save1=None
# save2=None
kf = KFold(len(self.x_train), n_folds=self.k_cross)
own_kappa = []
for train_idx, test_idx in kf:
# print train_idx,test_idx
# exit(0)
x_train, x_test = self.x_train[train_idx], self.x_train[test_idx]
y_train, y_test = self.y_train[train_idx], self.y_train[test_idx]
dim_red = LDA()
x_train = dim_red.fit_transform(x_train, y_train)
# with open('dumped_dim_red_'+str(i)+'.pkl', 'wb') as fid:
# cPickle.dump(dim_red, fid)
# with open('dumped_dim_red_'+str(i)+'.pkl', 'rb') as fid:
# dim_red=cPickle.load(fid)
x_test = dim_red.transform(x_test)
# with open('dumped_'+str(j)+'_'+str(i)+'.pkl', 'rb') as fid:
# stat_obj=cPickle.load(fid)
# x_train = dim_red.transform(x_train)
# x_test = dim_red.transform(x_test)
stat_obj = self.stat_class() # reflection bitches
stat_obj.train(x_train, y_train)
# with open('dumped_'+str(j)+'_'+str(i)+'.pkl', 'wb') as fid:
# cPickle.dump(stat_obj, fid)
# with open('dumped_'+str(j)+'_'+str(i)+'.pkl', 'rb') as fid:
# stat_obj=cPickle.load(fid)
y_pred = [0 for i in xrange(len(y_test))]
if (int(jk) == 1):
# print "test_idx"
save1 = stat_obj
save2 = dim_red
for i in range(len(x_test)):
# print len(x_test[i])
val = int(np.round(stat_obj.predict(x_test[i])))
if val > self.range_max: val = self.range_max
if val < self.range_min: val = self.range_min
y_pred[i] = [val]
y_pred = np.matrix(y_pred)
cohen_kappa_rating = own_wp.quadratic_weighted_kappa(
y_test, y_pred, self.range_min, self.range_max)
self.values.append(cohen_kappa_rating)
# print stat_obj.predict(x_train)
# linear_k_cross = k_fold_cross_validation(cross_valid_k,linear_regression,X_train,Y_train,range_min,range_max)
# linesar_accuracy.append(linear_k_cross.execute(i,0))
# logistic_k_cross = k_fold_cross_validation(cross_valid_k,logistic_regression,X_train,Y_train,range_min,range_max)
# logistic_accuracy.append(logistic_k_cross.execute(i,1))
# svr_k_cross = k_fold_cross_validation(cross_valid_k,support_vector_regression,X_train,Y_train,range_min,range_max)
# svr_accuracy.append(svr_k_cross.execute(i,2))
# svm_k_cross = k_fold_cross_validation(cross_valid_k,support_vector_machine,X_train,Y_train, range_min,range_max)
# svm_accuracy.append(svm_k_cross.execute(i,3))
return str(sum(self.values) / self.k_cross)
importance_w, Xw_cols = zip(*sorted(zip(importance_w, Xw_cols)))
fig = plt.figure(figsize=(6, 4), dpi=80).add_subplot(111)
plt.bar(range(len(Xw_cols)), importance_w, align='center')
plt.xticks(range(len(Xw_cols)), Xw_cols, rotation='vertical')
plt.xlabel('Features')
plt.ylabel('Importance of features')
plt.title("PCA for white wine")
plt.show()
# PCA for red wine
pca_transf_r = pca.fit_transform(Xr_minmax)
importance_r = pca.explained_variance_ratio_
print importance_r
importance_r, Xr_cols = zip(*sorted(zip(importance_r, Xr_cols)))
fig = plt.figure(figsize=(6, 4), dpi=80).add_subplot(111)
plt.bar(range(len(Xr_cols)), importance_r, color='red', align='center')
plt.xticks(range(len(Xr_cols)), Xr_cols, rotation='vertical')
plt.xlabel('Features')
plt.ylabel('Importance of features')
plt.title("PCA for red wine")
plt.show()
# LDA for white wine
from sklearn.lda import LDA
lda = LDA(n_components=None)
transf_lda = lda.fit_transform(Xw_minmax, Yw)
def fit_lda(df, active_features, y_col, k=2):
lda = LDA(n_components=k)
X = lda.fit_transform(df[active_features], df[y_col])
return X
plt.figure()
plt.pcolormesh(xx, yy, Z, cmap=cmap_light)
# Plot also the training points
plt.scatter(X[:, 0], X[:, 1], c=y, cmap=cmap_bold)
plt.axis('tight')
plt.axis('off')
plt.tight_layout()
plot_estimator(svc, X, y)
plt.show()
plt.clf()
###LDA###
sklearn_lda = LDA(n_components=2)
X_lda_sklearn = sklearn_lda.fit_transform(X, y)
print X_lda_sklearn
#4. Petal Length X Sepal Width for all 3
plt.scatter(iris.data[:, 1], iris.data[:, 2], c=iris.target)
plt.xlabel(iris.feature_names[1])
plt.ylabel(iris.feature_names[2])
plt.scatter(iris.data[0:150, 1], iris.data[0:150, 2], c=iris.target[0:150])
plt.xlabel(iris.feature_names[1])
plt.ylabel(iris.feature_names[2])
from sklearn import svm
svc = svm.SVC(kernel='linear', C=10) #The > C, the wider the margin between groups
from sklearn import datasets
X = X_lda_sklearn #reassigning to LDA
class DotProduct(DP1):
name = 'LDA'
LDA_components = 2
def __init__(self, X,Y, room, bin_size):
assert (room[0][1]-room[0][0])%bin_size == 0
assert (room[1][1]-room[1][0])%bin_size == 0
self.bin_size=bin_size
self.room=room
self.xblen = (room[0][1]-room[0][0])/bin_size
self.yblen = (room[1][1]-room[1][0])/bin_size
self.bins = self.xblen*self.yblen
self.labels = np.unique(Y)
newX = np.zeros([X.shape[0],self.LDA_components+self.bins])
newX[:,-self.bins:] = X[:,-self.bins:]
self.lda = LDA(n_components=self.LDA_components)
tmp = self.lda.fit_transform(X[:,:-self.bins],Y)
import pdb; pdb.set_trace()
newX[:,:self.LDA_components] = tmp
# This is if X = [cell1, cell2, ..., celln, binfrac1,...,binfrac k^2]
self.train(newX,Y,room, bin_size)
def classify(self,X):
bin_frac = X[-self.bins:].reshape([self.xblen,self.yblen])
X = X[:-self.bins]
X = np.squeeze(self.lda.transform(X))
#self.base[cell id, lbl, xbin, ybin] = rate
cntxt0 = np.einsum('cxy,c,xy',self.base[:,0,:,:],X,bin_frac)
cntxt1 = np.einsum('cxy,c,xy',self.base[:,1,:,:],X,bin_frac)
if logging.getLogger().level <= 5:
tmp0 = 0
for cell in range(len(X)):
tmp0 += np.sum(X[cell]*bin_frac*self.base[cell,0,:,:])
tmp1 = 0
for cell in range(len(X)):
tmp1 += np.sum(X[cell]*bin_frac*self.base[cell,1,:,:])
assert np.allclose(tmp0,cntxt0)
assert np.allclose(tmp1,cntxt1)
#import pdb; pdb.set_trace()
if cntxt0 > cntxt1:
return {self.labels[0]: 1,
self.labels[1]: 0}
else:
return {self.labels[0]: 0,
self.labels[1]: 1}
'''
# Normalize
if cntxt0 != 0 or cntxt1 != 0:
mag = cntxt0+cntxt1
else:
mag = 1
cntxt0 /= mag
cntxt1 /= mag
assert (round(cntxt0 + cntxt1,5) in [0,1])'''
return {self.labels[0]: cntxt0,
self.labels[1]: cntxt1}
plt.xlabel('LD 1')
plt.ylabel('LD 2')
plt.legend(loc='lower right')
# plt.tight_layout()
# plt.savefig('./figures/lda2.png', dpi=300)
plt.show()
#############################################################################
print(50 * '=')
print('Section: LDA via scikit-learn')
print(50 * '-')
lda = LDA(n_components=2)
X_train_lda = lda.fit_transform(X_train_std, y_train)
lr = LogisticRegression()
lr = lr.fit(X_train_lda, y_train)
plot_decision_regions(X_train_lda, y_train, classifier=lr)
plt.xlabel('LD 1')
plt.ylabel('LD 2')
plt.legend(loc='lower left')
# plt.tight_layout()
# plt.savefig('./images/lda3.png', dpi=300)
plt.show()
X_test_lda = lda.transform(X_test_std)
plot_decision_regions(X_test_lda, y_test, classifier=lr)
#Construct dictionary for saving array
final_data = {'M1_d_data':M1_d_data,'S1_d_data':S1_d_data,'pmd_d_data':pmd_d_data,'pmv_d_data':pmv_d_data,'M1_c_data':M1_c_data,'S1_c_data':S1_c_data,'pmd_c_data':pmd_c_data,'pmv_c_data':pmv_c_data,'targets':targets}
#Construct temparary dictionary for figure generation
final_data_no_targets = {'M1 at Delivery':M1_d_data,'S1 at Delivery':S1_d_data,'PmD at Delivery':pmd_d_data,'PmV at Delivery':pmv_d_data,'M1 at Cue':M1_c_data,'S1 at Cue':S1_c_data,'PmD at Cue':pmd_c_data,'PmV at Cue':pmv_c_data}
np.save("multi_reward"+filename[-15:-4]+"_hists_"+name_of_bin,(final_data,unit_names))
#Perform PCA on PSTH followed By LDA on PCA transform of PSTH data and save figure showing results for each bin
for key,value in final_data_no_targets.iteritems():
print key
lda = LDA(n_components=2)
pca = RandomizedPCA(n_components=20)
proj = pca.fit_transform(value)
proj = lda.fit_transform(proj,targets)
print proj.shape
plt.clf()
plt.scatter(proj[:, 0], proj[:, 1], c=targets)
plt.title(key+" from "+name_of_bin)
plt.xlabel("LD1")
plt.ylabel("LD2")
plt.colorbar()
plt.savefig(key+" from "+name_of_bin+"s.png")
plt.clf()
for line in fileLines:
result.append(line.split())
return result
## II.1.1 LDA using sklearn package ##
trainingData = fileParser("IrisTrain2014.dt") # Parse iris files
testData = fileParser("IrisTest2014.dt")
x = []
y = []
for entry in trainingData:
x.append([float(entry[0]), float(entry[1])])
y.append(int(entry[2]))
lda = LDA(n_components = 2) # initialize LDA for 2 parameter entries
transformedLda = lda.fit_transform(x,y)
# Function to plot the transformed data
def plotLda(lda):
colors = ["red", "green", "blue"]
i=0
while(i < 3):
xs = []
ys = []
n = 0
while(n < len(lda)):
if(y[n] == i):
xs.append(lda[n][0])
ys.append(lda[n][1])
n += 1
plotter.scatter(x=xs, y=ys, color = colors[i])
plt.scatter(Xred_pca[y==lab, 0],Xred_pca[y==lab, 1],label=lab,c=col)
plt.xlabel('Principal Component 1')
plt.ylabel('Principal Component 2')
leg = plt.legend(loc='upper right', fancybox=True)
plt.show()
####################################################
########## Parte d: LDA ############################
####################################################
sklearn_lda = LDA(n_components=2)
Xred_lda = sklearn_lda.fit_transform(X_std,y)
cmap = plt.cm.get_cmap('Set1')
mclasses=(1,2,3,4,5,6,7,8,9)
mcolors = [cmap(i) for i in np.linspace(0,1,10)]
plt.figure(figsize=(12, 8))
for lab, col in zip(mclasses,mcolors):
plt.scatter(Xred_lda[y==lab, 0],Xred_lda[y==lab, 1],label=lab,c=col)
plt.xlabel('LDA/Fisher Direction 1')
plt.ylabel('LDA/Fisher Direction 2')
leg = plt.legend(loc='upper right', fancybox=True)
plt.show()
#
# ####################################################
# ########## Parte f: Construcción Clasificador ######
first_half = np.vstack(first_half)
second_half = np.hstack(Data[no_mappings/2:no_mappings,:,:,:])
second_half = np.vstack(second_half)
Data = np.vstack([first_half,second_half])
# for true targets uncomment next line
#targets = np.hstack([targets,targets])
#for random targets uncomment next line
targets = np.random.randint(1,no_locations+1,no_mappings*no_locations*no_thwacks)
lda = LDA(n_components=14)
pca = RandomizedPCA(n_components = 125)
classifier = KNeighborsClassifier(8)
proj = pca.fit_transform(Data)
proj = lda.fit_transform(proj,targets)
proj1 = pca.fit_transform(Data)
proj1 = lda.fit_transform(proj1,mapping_targets)
print(file)
plt.clf()
plt.scatter(proj[0:proj.shape[0]/2,0],proj[0:proj.shape[0]/2,1],c=targets[0:targets.shape[0]/2])
plt.title(file.rsplit('_')[0]+'_'+file.rsplit('_')[1]+" Before "+file.rsplit('_')[2]+" injection")
plt.colorbar()
plt.ylabel("LD1")
plt.xlabel("LD2")
plt.savefig(file.rsplit('_')[0]+'_'+file.rsplit('_')[1]+" Before "+file.rsplit('_')[2]+file[-11:-4]+" injection.svg")
plt.show()
plt.clf()
plt.scatter(proj[proj.shape[0]/2:proj.shape[0],0],proj[proj.shape[0]/2:proj.shape[0],1],c=targets[targets.shape[0]/2:targets.shape[0]])
eigen_vals = pca.explained_variance_ratio_ # Egan-values for each PCAs (Importance) # --- Fitting Model with PCA pca = PCA(n_components=2) # Only take first 2 PCs lr = LogisticRegression() X_train_pca = pca.fit_transform(X_train_std) # fit with trainset X_test_pca = pca.transform(X_test_std) # only transform with testset lr.fit(X_train_pca, Y_train) # (2) Linear Discriminant Analsysi (LDA) - linear separatible # --- Evaluate Importance of LDA from sklearn.linear_model import LogisticRegression from sklearn.lda import LDA lda = LDA(n_components=None) X_train_lda = lda.fit_transform(X_train_std, Y_train) # fit with trainset, (x, y) supervized eigen_vals = lda.explained_variance_ratio_ # Egan-values for each LDAs (Importance) # --- Fitting Model with LDA lda = LDA(n_components=2) X_train_lda = lda.fit_transform(X_train_std, Y_train) # fit with trainset, (x, y) supervized X_test_lda = lda.transform(X_test_std) # only transform with testset lr.fit(X_train_lda, Y_train) # (3) Kernel Principal Component Analysis (K-PCA) - non-linear separatible from sklearn.decomposition import KernelPCA scikit_kpca = KernelPCA(n_components=2, kernel='rbf', gamma=15) # can choose other kernel methods, 2 PCAs = features X_skernpca = scikit_kpca.fit_transform(X_train_std)
from sklearn.svm import LinearSVC
kappa_scorer = make_scorer(cohen_kappa_score)
grid = GridSearchCV(LinearSVC(), param_grid={'C': [1, 10]}, scoring=kappa_scorer)
#EXEMPLO 20 - LATENT FACTOR ANALYSIS (LFA)
from sklearn.decomposition import FactorAnalysis
fact_2c = FactorAnalysis(n_components=2)
X_factor = fact_2c.fit_transform(iris.data)
plt.scatter(X_factor[:,0], X_factor[:,1], c=iris.target,
alpha=0.8, s=60, marker='o', edgecolors='white')
plt.show()
#EXEMPLO 21 - LINEAR DISCRIMINANT ANALYSIS (LDA)
from sklearn.lda import LDA
lda_2c = LDA(n_components=2)
X_lda_2c = lda_2c.fit_transform(iris.data, iris.target)
plt.scatter(X_lda_2c[:,0], X_lda_2c[:,1], c=iris.target,
alpha=0.8, edgecolors='none'); plt.show()
#EXEMPLO 22 - KERNEL PCA
from sklearn.decomposition import KernelPCA
kpca_2c = KernelPCA(n_components=2, kernel='rbf')
X_kpca_2c = kpca_2c.fit_transform(fake_circular_data)
plt.scatter(X_kpca_2c[:,0], X_kpca_2c[:,1], c=fake_circular_target,
alpha=0.8, s=60, marker='o', edgecolors='white')
plt.show()
#EXEMPLO 22 - MARS
import numpy
from pyearth import Earth
from matplotlib import pyplot
plt.plot(transformedXg[:x1g.shape[0], 0], transformedXg[:x1g.shape[0], 1], 'o') plt.plot(transformedXg[x1g.shape[0]:, 0], transformedXg[x1g.shape[0]:, 1], 'x') plt.show() # <h2 style="color:purple"> LDA </h2> # In[25]: from sklearn.lda import LDA lda = LDA(n_components=5) xg = pd.concat([x1g, x2g]) yg = np.zeros(xg.shape[0]) yg[:x1g.shape[0]] = np.ones(x1g.shape[0]) transformedXg2 = lda.fit_transform(xg, yg) print transformedXg2.shape plt.plot(transformedXg2[:x1g.shape[0], 0], np.zeros(x1g.shape[0]), 'o') plt.plot(transformedXg2[x1g.shape[0]:, 0], np.zeros(x2g.shape[0]), 'x') plt.show() # <h2 style="color:purple"> Supervised Graph</h2> # In[ ]: # <h2 style="color:purple"> Unsupervised Graph</h2> # In[ ]: # <h2 style="color:green"> Classify </h2>
class DotProduct(DP1):
name = 'LDA'
LDA_components = 2
def __init__(self, X, Y, room, bin_size):
assert (room[0][1] - room[0][0]) % bin_size == 0
assert (room[1][1] - room[1][0]) % bin_size == 0
self.bin_size = bin_size
self.room = room
self.xblen = (room[0][1] - room[0][0]) / bin_size
self.yblen = (room[1][1] - room[1][0]) / bin_size
self.bins = self.xblen * self.yblen
self.labels = np.unique(Y)
newX = np.zeros([X.shape[0], self.LDA_components + self.bins])
newX[:, -self.bins:] = X[:, -self.bins:]
self.lda = LDA(n_components=self.LDA_components)
tmp = self.lda.fit_transform(X[:, :-self.bins], Y)
import pdb
pdb.set_trace()
newX[:, :self.LDA_components] = tmp
# This is if X = [cell1, cell2, ..., celln, binfrac1,...,binfrac k^2]
self.train(newX, Y, room, bin_size)
def classify(self, X):
bin_frac = X[-self.bins:].reshape([self.xblen, self.yblen])
X = X[:-self.bins]
X = np.squeeze(self.lda.transform(X))
#self.base[cell id, lbl, xbin, ybin] = rate
cntxt0 = np.einsum('cxy,c,xy', self.base[:, 0, :, :], X, bin_frac)
cntxt1 = np.einsum('cxy,c,xy', self.base[:, 1, :, :], X, bin_frac)
if logging.getLogger().level <= 5:
tmp0 = 0
for cell in range(len(X)):
tmp0 += np.sum(X[cell] * bin_frac * self.base[cell, 0, :, :])
tmp1 = 0
for cell in range(len(X)):
tmp1 += np.sum(X[cell] * bin_frac * self.base[cell, 1, :, :])
assert np.allclose(tmp0, cntxt0)
assert np.allclose(tmp1, cntxt1)
#import pdb; pdb.set_trace()
if cntxt0 > cntxt1:
return {self.labels[0]: 1, self.labels[1]: 0}
else:
return {self.labels[0]: 0, self.labels[1]: 1}
'''
# Normalize
if cntxt0 != 0 or cntxt1 != 0:
mag = cntxt0+cntxt1
else:
mag = 1
cntxt0 /= mag
cntxt1 /= mag
assert (round(cntxt0 + cntxt1,5) in [0,1])'''
return {self.labels[0]: cntxt0, self.labels[1]: cntxt1}
labels = set(df.columns.values)
labels.remove('y')
X_raw = df[list(labels)]
X_train, _, _ = one_hot_dataframe(X_raw, [
'job', 'marital', 'education', 'default', 'housing', 'loan', 'contact',
'month', 'poutcome'
],
replace=True)
y_train = [1 if i == 'yes' else 0 for i in df.y]
reductions = []
pca = PCA(n_components=2)
reductions.append(pca.fit_transform(X_train, y_train))
lda = LDA(n_components=2)
reductions.append(lda.fit_transform(X_train, y_train))
isomap = Isomap(n_components=2)
reductions.append(isomap.fit_transform(X_train, y_train))
lle = LocallyLinearEmbedding(n_components=2, method='standard')
reductions.append(lle.fit_transform(X_train, y_train))
for reduced_X in reductions:
plt.figure()
red_x = []
red_y = []
blue_x = []
blue_y = []
green_x = []
green_y = []
for i in range(len(reduced_X)):
n1 = Y[Y==0].shape[0]
n2 = Y[Y==1].shape[0]
print "PCA:"
pca = PCA(n_components=Components)
transformed = pca.fit_transform(X)
plt.plot(transformed[:n1,0],transformed[:n1,1],'o')
plt.plot(transformed[n1:,0],transformed[n1:,1],'x')
plt.show()
print "LDA:"
lda = LDA()
transformed2 = lda.fit_transform(X,Y)
plt.plot(transformed2[:n1,0],np.zeros(n1),'o')
plt.plot(transformed2[n1:,0],np.zeros(n2),'x')
plt.show()
print "unsupervised:"
numFeature = X.shape[1]
numData = X.shape[0]
numNode = numFeature + numData
A = np.zeros((numNode,numNode))
# construct feature-data
for i in range(numData):
for j in xrange(numFeature):
A[i+numFeature,j] = X.iloc[i,j]
A[j,i+numFeature] = X.iloc[i,j]
# remove axis spines
ax.spines["top"].set_visible(False)
ax.spines["right"].set_visible(False)
ax.spines["bottom"].set_visible(False)
ax.spines["left"].set_visible(False)
plt.grid()
plt.tight_layout
plt.show()
os.chdir("F:\Analytics\ISB Study\Capstone\dir_data\dir_data")
X_train, y_train, X_test, y_test, X_val, y_val = load_svmlight_files(("train\\vision_cuboids_histogram.txt", "test\\vision_cuboids_histogram.txt","validation\\vision_cuboids_histogram.txt"))
np.unique(y_train)
sklearn_lda = LDA(n_components=30)
X_lda_sklearn = sklearn_lda.fit_transform(X_train.todense(), y_train)
plot_scikit_lda(X_lda_sklearn, title='LDA vision_cuboids_histogram')
# PCA
sklearn_pca = sklearnPCA(n_components=30)
X_pca = sklearn_pca.fit_transform(X_train.todense())
plot_pca(title = 'PCA vision_cuboids_histogram')
#
X_ldapca_sklearn = sklearn_pca.fit_transform(X_lda_sklearn)
plot_scikit_lda(X_ldapca_sklearn, title='LDA+PCA LDA vision_cuboids_histogram', mirror=(-1))
x_test = sc.transform(x_test)
import numpy as np
cov_mat = np.cov(x_train.T)
eigen_vals, eigen_vecs = np.linalg.eig(cov_mat)
print('eigenvals', eigen_vals)
tot = sum(eigen_vals)
var_exp = [(i / tot) for i in sorted(eigen_vals, reverse=True)]
cum_var_exp = np.cumsum(var_exp)
import matplotlib.pyplot as plt
plt.figure(figsize=(6, 4))
plt.bar(range(1, 14),
var_exp,
alpha=0.5,
align='center',
label='ind explained variance')
plt.step(range(1, 14),
cum_var_exp,
where='mid',
label='cum explained variance')
plt.ylabel('Explained Variance')
plt.xlabel('Principle Components')
plt.legend(loc=0)
plt.show()
from sklearn.lda import LDA
lda = LDA(n_components=2)
x_train_lda = lda.fit_transform(x_train, y_train)
final_data_no_targets = {
'M1 at Delivery': M1_d_data,
'S1 at Delivery': S1_d_data,
'PmD at Delivery': pmd_d_data,
'PmV at Delivery': pmv_d_data,
'M1 at Cue': M1_c_data,
'S1 at Cue': S1_c_data,
'PmD at Cue': pmd_c_data,
'PmV at Cue': pmv_c_data
}
np.save("multi_reward" + filename[-15:-4] + "_hists_" + name_of_bin,
(final_data, unit_names))
#Perform PCA on PSTH followed By LDA on PCA transform of PSTH data and save figure showing results for each bin
for key, value in final_data_no_targets.iteritems():
print key
lda = LDA(n_components=2)
pca = RandomizedPCA(n_components=20)
proj = pca.fit_transform(value)
proj = lda.fit_transform(proj, targets)
print proj.shape
plt.clf()
plt.scatter(proj[:, 0], proj[:, 1], c=targets)
plt.title(key + " from " + name_of_bin)
plt.xlabel("LD1")
plt.ylabel("LD2")
plt.colorbar()
plt.savefig(key + " from " + name_of_bin + "s.png")
plt.clf()
def lda(X, y, components=10):
lda = LDA(n_components=components)
return lda.fit_transform(X, y)
np.mean(score)
#0.913
#Convert array to dataframe
df = pd.DataFrame(x, columns = ['C1', 'C2', 'C3'])
#3 dimentional plot of principal components (eigenvectors)
fig = plt.figure(1)
ax = fig.add_subplot(111, projection = '3d')
ax.scatter(df['C1'].tolist(), df['C2'].tolist(), df['C3'].tolist(), c = 'b', marker = '*')
ax.set_xlabel('C1')
ax.set_ylabel('C2')
ax.set_zlabel('C3')
ax.legend()
#Linear discriminant analysis
#Standardize
da = LDA(n_components = 3)
#Fit model and transform data
a = da.fit_transform(X, Y)
#Perform KNN algorithm
neigh1 = KNeighborsClassifier(n_neighbors = 3)
neigh1.fit(a, Y)
#Get cross validation metrics
#Create validation set with k-fold cross validation
#Test for accuracy of validation set
score1 = cross_val_score(neigh1, a, Y, scoring = 'accuracy', cv = 10)
np.mean(score1)
#0.973
def execute(self,i,j):
global save1
global save2
jk=i
# print type(jk)
# dim_red = LDA()
# dim_red.fit_transform(self.x_train, self.y_train)
# with open('dumped_dim_red_'+str(i)+'.pkl', 'wb') as fid:
# cPickle.dump(dim_red, fid)
# x_train = dim_red.transform(self.x_train)
# x_test = dim_red.transform(self.y_train)
# stat_obj = self.stat_class() # reflection bitches
# stat_obj.train(x_train, x_test)
# print len(x_train)
# with open('dumped_'+str(j)+'_'+str(i)+'.pkl', 'wb') as fid:
# cPickle.dump(stat_obj, fid)
# save1=None
# save2=None
kf = KFold(len(self.x_train), n_folds=self.k_cross)
own_kappa = []
for train_idx, test_idx in kf:
# print train_idx,test_idx
# exit(0)
x_train, x_test = self.x_train[train_idx], self.x_train[test_idx]
y_train, y_test = self.y_train[train_idx], self.y_train[test_idx]
dim_red = LDA()
x_train = dim_red.fit_transform(x_train, y_train)
# with open('dumped_dim_red_'+str(i)+'.pkl', 'wb') as fid:
# cPickle.dump(dim_red, fid)
# with open('dumped_dim_red_'+str(i)+'.pkl', 'rb') as fid:
# dim_red=cPickle.load(fid)
x_test = dim_red.transform(x_test)
# with open('dumped_'+str(j)+'_'+str(i)+'.pkl', 'rb') as fid:
# stat_obj=cPickle.load(fid)
# x_train = dim_red.transform(x_train)
# x_test = dim_red.transform(x_test)
stat_obj = self.stat_class() # reflection bitches
stat_obj.train(x_train,y_train)
# with open('dumped_'+str(j)+'_'+str(i)+'.pkl', 'wb') as fid:
# cPickle.dump(stat_obj, fid)
# with open('dumped_'+str(j)+'_'+str(i)+'.pkl', 'rb') as fid:
# stat_obj=cPickle.load(fid)
y_pred = [ 0 for i in xrange(len(y_test)) ]
if(int(jk)==1):
# print "test_idx"
save1=stat_obj
save2=dim_red
for i in range(len(x_test)):
# print len(x_test[i])
val = int(np.round(stat_obj.predict(x_test[i])))
if val > self.range_max: val = self.range_max
if val < self.range_min: val = self.range_min
y_pred[i] = [val]
y_pred = np.matrix(y_pred)
cohen_kappa_rating = own_wp.quadratic_weighted_kappa(y_test,y_pred,self.range_min,self.range_max)
self.values.append(cohen_kappa_rating)
# print stat_obj.predict(x_train)
# linear_k_cross = k_fold_cross_validation(cross_valid_k,linear_regression,X_train,Y_train,range_min,range_max)
# linesar_accuracy.append(linear_k_cross.execute(i,0))
# logistic_k_cross = k_fold_cross_validation(cross_valid_k,logistic_regression,X_train,Y_train,range_min,range_max)
# logistic_accuracy.append(logistic_k_cross.execute(i,1))
# svr_k_cross = k_fold_cross_validation(cross_valid_k,support_vector_regression,X_train,Y_train,range_min,range_max)
# svr_accuracy.append(svr_k_cross.execute(i,2))
# svm_k_cross = k_fold_cross_validation(cross_valid_k,support_vector_machine,X_train,Y_train, range_min,range_max)
# svm_accuracy.append(svm_k_cross.execute(i,3))
return str(sum(self.values)/self.k_cross)
def FLD_r(X, y):
# can change n_components to desired value
fld = LDA()
return fld.fit_transform(X, y)
# A cada dato se le asigna la clase y un color para diferenciarlas entre si
for lab, col in zip(mclasses,mcolors):
plt.scatter(Xred_pca[y==lab, 0],Xred_pca[y==lab, 1],label=lab,c=col)
# Se configuran las etiquetas
plt.xlabel('Principal Component 1')
plt.ylabel('Principal Component 2')
leg = plt.legend(loc='upper right', fancybox=True)
######## Pregunta (d) ############################################################
# Se utiliza LDA con dos dimensiones
sklearn_lda = LDA(n_components=2)
# Se ajusta a los datos de entrenamiento
Xred_lda = sklearn_lda.fit_transform(X_std,y)
# Se escoge la paleta de colores
cmap = plt.cm.get_cmap('hsv')
# Se definen las clases
mclasses=(1,2,3,4,5,6,7,8,9)
mcolors = [cmap(i) for i in np.linspace(0,1,10)]
# Se establece el tamanno de la figura
plt.figure(figsize=(12, 8))
# A cada dato se le asigna la clase y un color para diferenciarlas entre si
for lab, col in zip(mclasses,mcolors):
plt.scatter(Xred_lda[y==lab, 0],Xred_lda[y==lab, 1],label=lab,c=col)
def lda_projected_X(X, y, n):
pca = LDA(n_components=n)
return pca.fit_transform(X, y)
# Project the tranformed points
w_lda = np.dot(np.linalg.inv(s_within), mean_1 - mean_2)
w_lda = w_lda / np.linalg.norm(w_lda)
label_1_projected = np.dot(label_1_transformed, w_lda)
label_2_projected = np.dot(label_2_transformed, w_lda)
# plot the transformed points
plt.plot(label_1_projected, np.zeros_like(label_1_projected), 'o', c='r')
plt.plot(label_2_projected, np.zeros_like(label_2_projected), 'o', c='b')
plt.show()
# LDA using sklearn
from sklearn.lda import LDA
points = np.concatenate((label_1, label_2), axis=0)
label = np.concatenate((np.zeros((20, 1)), np.zeros((20, 1)) + 1))
lda = LDA()
skl_transform = lda.fit_transform(points, label)
f, ax = plt.subplots(1, 2)
ax[0].plot(skl_transform[:20], np.zeros_like(label_1_projected), 'o', c='r')
ax[0].plot(skl_transform[20:], np.zeros_like(label_2_projected), 'o', c='b')
ax[0].set_title('SKlearn Projection')
ax[1].plot(label_1_projected, np.zeros_like(label_1_projected), 'o', c='r')
ax[1].plot(label_2_projected, np.zeros_like(label_2_projected), 'o', c='b')
ax[1].set_title('Manual Projection')
plt.show()
plt.title('Projection onto first 2 PC space')
plt.xlabel('Principle Component 1')
plt.ylabel('Principle Component 2')
fig1.savefig('./Plots/2_PCA1.png')
# Generate scatter plot on the second 2 PC spacefig2 = plt.figure()
fig2 = plt.figure(figsize=(16.0, 9.0))
plt.scatter( x=pc_data[:,2], y=pc_data[:,3], c=[color_dict[c] for c in df.iloc[:,0]] )
plt.title('Projection onto second 2 PC space')
plt.xlabel('Principle Component 3')
plt.ylabel('Principle Component 4')
fig2.savefig('./Plots/2_PCA2.png')
# Apply LDA to project PC space onto first 2 LD space
lda = LDA()
ld_data = lda.fit_transform(X=pc_data, y=df.iloc[:,0])
fig3 = plt.figure(figsize=(16.0, 9.0))
plt.scatter( x=ld_data[:,0], y=ld_data[:,1], c=[color_dict[c] for c in df.iloc[:,0]] )
plt.title('Projection onto first 2 LD space')
plt.xlabel('Linear Discriminant 1')
plt.ylabel('Linear Discriminant 2')
fig3.savefig('./Plots/2_LDA1.png')
# Apply LDA to project PC space onto second 2 LD space
fig4 = plt.figure(figsize=(16.0, 9.0))
plt.scatter( x=ld_data[:,2], y=ld_data[:,3], c=[color_dict[c] for c in df.iloc[:,0]] )
plt.title('Projection onto second 2 LD space')
plt.xlabel('Linear Discriminant 3')
plt.ylabel('Linear Discriminant 4')
fig4.savefig('./Plots/2_LDA2.png')
train.head() # In[ ]: lda = LDA(n_components=2) pca = PCA(n_components=2) # In[ ]: scaler = StandardScaler() trains = scaler.fit_transform(train) trans = pca.fit_transform(trains, labels) dfp = pd.DataFrame(trans, index=train.index) trans = lda.fit_transform(trains, labels) df = pd.DataFrame(trans, columns=["lda"], index=train.index) # In[ ]: df["labels"] = labels dfp["labels"] = labels # In[ ]: df.plot(kind='scatter', x='lda', y='labels') # In[ ]: dfp.plot(kind='scatter', x='pca1', y='pca2', c="labels")
