def __init__(self):
train_data, train_label, test_data, test_label = read_data(2)
train_data.astype('float16')
train_data = train_data / 255
test_data.astype('float16')
test_data = test_data / 255
self.train_data = dataset(train_data)
self.train_label = dataset(self.modify_label(train_label))
self.test_data = dataset(test_data)
self.test_label = dataset(self.modify_label(test_label))
def main():
train_data, train_label, test_data, test_label = read_data("HOG")
#load_predict("mlp3", test_data, test_label)
print("Fitting...")
clf = SVM(train_data, train_label)
save_model(clf, "svm7")
print("Predicting...")
#pca = PCA(n_components=100)
#test_data = pca.fit_transform(test_data)
#test_data = test_data[:300]
#test_label = test_label[:300]
predict_label = clf.predict(test_data)
#predict_label = m3(test_data)
acc = accuracy(test_label, predict_label)
print("Accuracy: ", acc)
import numpy as np
import sys
import chainer
from chainer import Variable, serializers
from prepare import get_typemap, read_data
from model import create_model2
inputdata_file = sys.argv[1]
model_file_name = sys.argv[2]
state_file_name = sys.argv[3]
typemap = get_typemap(inputdata_file)
x, t = read_data(inputdata_file, typemap)
dim = [x.shape[1], 120, 50, 1]
model, optimizer = create_model2(dim)
serializers.load_npz(model_file_name, model)
serializers.load_npz(state_file_name, optimizer)
y = model.fwd(Variable(x)).data
print("expect actual diff acc")
for r in np.hstack([t, y, y - t, 1 - abs((y - t) / t)]):
print("%6d %6d %5d %.2f" % (r[0], r[1], r[2], r[3]))
ac = (1 - abs(y - t) / t).mean()
print("total acc = %.3f" % ac)
d = 1 - abs(y - t) / t
d.sort(axis=0)
ac = d[:50].mean()
print("total acc = %.3f" % ac)
# -*- coding: utf-8 -*-
from prepare import read_data, save_data, prepare_email
from prepare import validate_email, validate_password
if __name__ == "__main__":
start = 35
data = read_data('data_%s00000_%s00000.txt' % (start, start + 2), sep=" ")
counter = 0
good_counter = 0
# http://technet.microsoft.com/en-us/library/gg524800.aspx
with open('cleansed.txt', 'w') as cleansed, \
open('dirty.txt', 'w') as dirty:
for d in data:
counter += 1
if len(d) == 1:
pass
# ('*****@*****.**', '50537580')
elif len(d) == 2:
email, password = d
if validate_password(password):
email = prepare_email(email)
if validate_email(email):
cleansed.write("%s\t%s\n" % (email, password))
good_counter += 1
# ('597305222', '*****@*****.**', '13945560329')
elif len(d) == 3 and d[0] in d[1]:
email, password = d[1], d[2]
markers = ["^", "o", "*", ".", ",", "v", ">", "<", "+", "1", "2"]
colors = ["b", "g", "r", "c", "m", "y", "k", "w", "teal", "darkred", "indigo"]
for i in range(num_of_hidden_layers):
index = np.where(y == i)
plt.scatter(lo[index], la[index], marker=markers[i], color=colors[i])
# parameters
mode = sys.argv[1]
datafile = sys.argv[2]
background_image = None
if len(sys.argv) > 3: background_image = sys.argv[3]
outputfile = None
if len(sys.argv) > 4: outputfile = sys.argv[4]
# read csv data
lo, la, data = read_data(datafile)
# plot
min_x = math.floor(min(lo) * 10) / 10
max_x = math.ceil(max(lo) * 10) / 10
min_y = math.floor(min(la) * 10) / 10
max_y = math.ceil(max(la) * 10) / 10
x_ticks = np.arange(min_x, max_x, 0.2) # x label
y_ticks = np.arange(min_y, max_y, 0.2) # y label
extent = [min_x, max_x, min_y, max_y] # image size
plt.figure(figsize=(18,10)) # plot area size
subplot_rows = 2
subplot_cols = 4
num_of_hidden_layers_list = [2, 3, 4, 5, 6, 7, 8, 9]