def run_train(algorithm):
X, y = code.loaddata()
validate_idx = code.make_random_indices(200, len(y))
estimate_idx = code.make_random_indices(1000, len(y))
Parameters = {'algorithm': algorithm}
return code.TrainMyClassifier(X[estimate_idx], X[validate_idx], Parameters,
y[estimate_idx])
def cross_validate_pca(algorithm):
import sklearn.decomposition
X, y = code.loaddata()
#pca = sklearn.decomposition.PCA(n_components='mle', svd_solver='full')
pca = sklearn.decomposition.PCA(n_components=6)
X = pca.fit_transform(X)
return code.MyCrossValidate(X, 5, {'algorithm': algorithm}, y)
def run_cross_validation(algorithm, length=5000):
X, y = code.loaddata()
if length > 0:
indices = code.make_random_indices(length, len(y))
X = X[indices]
y = y[indices]
return code.MyCrossValidate(X, 5, {
'algorithm': algorithm,
'ifprint': True
}, y)
def test_SVC():
import scipy.io as sio
import numpy as np
from sklearn.model_selection import KFold
from sklearn.cross_validation import train_test_split
import sklearn.svm as svm
from sklearn.metrics import confusion_matrix
from sklearn.metrics import classification_report
import random
# features = sio.loadmat("Proj2FeatVecsSet1.mat").get("Proj2FeatVecsSet1")
# targetOutput = sio.loadmat("Proj2TargetOutputsSet1.mat").get("Proj2TargetOutputsSet1")
#read in data
[features, targetOutput] = code.loaddata()
# print features, features.shape
# print targetOutput, targetOutput.shape
index = np.arange(features.shape[0])
#index = np.arange(10)
random.shuffle(index)
#print index, features.shape[0], index[0:3]
# print features[index[0:3]]
# print features[index[0:3]].shape
index_train = index[0:20000]
index_test = index[20000:]
features_train = features[index_train]
features_test = features[index_test]
targetOutput_train = targetOutput[index_train]
targetOutput_test = targetOutput[index_test]
#X_estimate, X_validate, y_estimate, y_validate = train_test_split(features_train,
# targetOutput_train, test_size=0.2, random_state=42)
param = ["linear", 2.0]
kf = KFold(n_splits=5, shuffle=True, random_state=17)
score = 0.0
score1 = 0.0
for train_index, test_index in kf.split(features_train):
X_estimate, X_validate = features_train[train_index], features_train[
test_index]
y_estimate, y_validate = targetOutput_train[
train_index], targetOutput_train[test_index]
estParam = code.SVM.TrainMyClassifierSVM(X_estimate, y_estimate)
score1 = estParam.score(X_validate, y_validate)
score += score1
print score1
print "average score: ", score / 5.0
y_predict_clf2 = code.SVM.TestMyClassifierSVM(features_test, estParam)
confMat = confusion_matrix(targetOutput_test, y_predict_clf2)
print confMat
print(classification_report(targetOutput_test, y_predict_clf2))
print estParam.score(features_test, targetOutput_test)
def print_confusion_matrices(algorithm, length=5000):
X, y = code.loaddata()
if length > 0:
indices = code.make_random_indices(length, len(y))
X = X[indices]
y = y[indices]
ytrain, clfs, conf_mats, conf_mat = code.MyCrossValidate(
X, 5, {'algorithm': algorithm}, y)
output = ''
newline = '\n'
for k in range(len(conf_mats)):
output += 'confusion matrix ' + str(k) + newline
output += code.print_confusion_matrix(conf_mats[k]) + newline
output += 'overall confusion matrix' + newline
output += code.print_confusion_matrix(conf_mat)
f = open('conf_mats_' + algorithm + '.txt', 'w')
f.write(output)
f.close()
def call_MyConfMatrix(algorithm, length=5000):
X, y = code.loaddata()
if length > 0:
indices = code.make_random_indices(length, len(y))
X = X[indices]
y = y[indices]