def main(): DATA_DIR = 'arcene' data = extract(os.path.join(DATA_DIR,'arcene_train.data')) data = data.T data, mean = centre(data) data_valid = extract(os.path.join(DATA_DIR,'arcene_train.data')) data_valid,mean = centre(data_valid.T, mean) labels = gen_labels(os.path.join(DATA_DIR, 'arcene_train.labels')) sig = get_sigma(data, labels) eig_val, eig_vec = gram(rbf, data, sig) alpha = get_lda(data, labels, rbf, sig) reduced_points = kernel_pca(rbf, 10, eig_vec, eig_val, data, data, sig) red_points = lda(alpha, data, data, rbf, sig)
def main():
DATA_DIR = 'arcene'
header = "arcene_.data"
index = header.find('.')
files = [header[:index] + x + header[index:] for x in ["train"]]
data = []
for f in files:
data = extract(os.path.join(DATA_DIR ,f))
lda(data)
def main():
k = 10
DATA_DIR = 'arcene'
train_file_labels = os.path.join(DATA_DIR, 'arcene_train.labels')
#data = extract(filename)
header = "arcene_train.data_10"
header2 = "arcene_.data"
index = header2.find('.')
train_labels = gen_labels(train_file_labels)
actual_files = ["test", "train", "valid"]
transformed_pca_data = []
data = []
data_files = [header2[:index] + x + header2[index:] for x in actual_files]
weight_files = [header + j for j in ['', '0', '00' ]]
weights = np.asarray(map(np.loadtxt, weight_files))
assert(weights[0].shape[1] == 10000 and weights[0].shape[0] == 10)
assert(weights[1].shape[1] == 10000 and weights[1].shape[0] == 100)
assert(weights[2].shape[1] == 10000 and weights[2].shape[0] == 1000)
for file in data_files:
data.append(extract(os.path.join(DATA_DIR,file)))
for i in xrange(len(data)):
for j in xrange(len(weights)):
transformed_pca_data.append(np.dot(weights[j], data[i].T))
#transformed_pca_data = np.asarray(map(np.loadtxt, files))
transformed_pca_train_data = np.asarray([transformed_pca_data[3], transformed_pca_data[4], transformed_pca_data[5]])
assert(len(transformed_pca_data) == 9)
for i in xrange(9):
if(i/3 == 0):
assert(transformed_pca_data[i].shape[1] == 700)
else:
assert(transformed_pca_data[i].shape[1] == 100)
if(i%3 == 0):
assert(transformed_pca_data[i].shape[0] == 10)
elif(i%3 == 1):
assert(transformed_pca_data[i].shape[0] == 100)
else:
assert(transformed_pca_data[i].shape[0] == 1000)
prior, means, cov, classes = train_data(train_labels, transformed_pca_train_data)
assert(len(means) == len(transformed_pca_train_data))
assert(len(cov) == len(transformed_pca_train_data))
#labels = gaussian_classify(transformed_pca_train_data[1], prior, means[1], cov[1], classes)
labels = naive_gaussian(transformed_pca_data[7], prior, means[1], cov[1], classes)
