for (x_string, label_string) in zip(fp_data, fp_labels):
x_string = x_string.strip()
label_string = label_string.strip()
if not x_string:
assert not label_string
continue
label = int(label_string)
if label not in (-1, 1):
logging.error("Unknown label: %d" % label)
sys.exit(2)
x_original = np.fromstring(x_string, sep=' ')
# Transform the features.
x = transform(x_original).flatten()
if not x.shape == weights.shape:
logging.error("Shapes of weight vector and transformed "
"data don't match")
sys.exit(3)
if np.abs(np.sum(weights)) < 1e-10:
logging.error("Zero vector provided")
sys.exit(4)
if label*np.inner(weights, x) >= 0:
accuracy += 1
total += 1
print("%d" % (total))
print("%d" % (accuracy))
# So we can use the same feature vectors.
from mapper import transform
if __name__ == "__main__":
if not len(sys.argv) == 3:
logging.error("Usage: evaluate.py weights.txt test_data.txt")
sys.exit(1)
with open(sys.argv[1], "r") as fp_weights:
weights = np.genfromtxt(fp_weights).flatten()
with open(sys.argv[2], "r") as fp_data:
#with open(sys.argv[3], "r") as fp_labels:
for x_string in fp_data:
x_string = x_string.strip()
x_original = np.fromstring(x_string, sep=' ')
x = transform(x_original).flatten() # Use our features.
if not x.shape == weights.shape:
logging.error("Shapes of weight vector and transformed "
"data don't match")
sys.exit(3)
#if (np.inner(weights, x) < 0):
# print -1
#else:
# print 1
print(-1 if np.inner(weights, x) < 0 else 1)
with open(sys.argv[2], "r") as fp_data:
with open(sys.argv[3], "r") as fp_labels:
for (x_string, label_string) in zip(fp_data, fp_labels):
x_string = x_string.strip()
label_string = label_string.strip()
if not x_string:
assert not label_string
continue
label = int(label_string)
if label not in (-1, 1):
logging.error("Unknown label: %d" % label)
sys.exit(2)
x_original = np.fromstring(x_string, sep=' ')
x = transform(x_original).flatten() # Use our features.
if not x.shape == weights.shape:
logging.error("Shapes of weight vector and transformed "
"data don't match")
sys.exit(3)
# if (np.inner(weights, x) < 0):
# print -1
# else:
# print 1
if label * np.inner(weights, x) >= 0:
accuracy += 1
total += 1
print("%f" % (float(accuracy) / total))
import sys
import numpy as np
sys.path.append(sys.argv[4])
from mapper import transform
if __name__ == "__main__":
with open(sys.argv[1], "r") as weights_file:
weights = np.genfromtxt(weights_file).flatten()
with open(sys.argv[2], "r") as data_file:
with open(sys.argv[3], "w") as pred_file:
for x_string in data_file:
# Apply the transformation for x and then continue.
# Compute the label and write it to the prediction file.
x_o = np.fromstring(x_string, sep=' ')
x_t = transform(x_o)
print("WEIGHTS: %s"%(weights.shape))
print("VALUES: %s"%(x_t.shape))
if np.inner(weights, x_t) > 0:
pred_file.write("-1\n")
else:
pred_file.write("+1\n")
import sys
import numpy as np
sys.path.append(sys.argv[4])
from mapper import transform
if __name__ == "__main__":
with open(sys.argv[1], "r") as weights_file:
weights = np.genfromtxt(weights_file).flatten()
with open(sys.argv[2], "r") as data_file:
with open(sys.argv[3], "w") as pred_file:
for x_string in data_file:
# Apply the transformation for x and then continue.
# Compute the label and write it to the prediction file.
x_o = np.fromstring(x_string, sep=' ')
x_t = transform(x_o)
print("WEIGHTS: %s" % (weights.shape))
print("VALUES: %s" % (x_t.shape))
if np.inner(weights, x_t) > 0:
pred_file.write("-1\n")
else:
pred_file.write("+1\n")
