def learning(weight):
for line in iter(sys.stdin.readline, ""):
label, sent = get_label_and_sentence(line)
phi = cf.create_features(sent)
pre_label = po.predict_one(weight, phi)
if int(pre_label) != int(label):
update_weights(weight, phi, int(label))
def learning(weight):
for line in iter(sys.stdin.readline, ""):
label, sent = get_label_and_sentence(line)
phi = cf.create_features(sent)
pre_label = po.predict_one(weight, phi)
if int(pre_label) != int(label):
update_weights(weight, phi, int(label))
def average_learning(weight):
updates = 0.0
average = defaultdict(lambda: 0.0)
for line in iter(sys.stdin.readline, ""):
label, sent = get_label_and_sentence(line)
phi = cf.create_features(sent)
pre_label = po.predict_one(weight, phi)
if int(pre_label) != int(label):
update_weights(weight, phi, int(label))
updates += 1.0
for key, value in weight.items():
average[key] = (average[key] * (updates - 1.0) + value) / updates
# copy to weight from average
for key in weight.keys():
weight[key] = average[key]
def average_learning(weight):
updates = 0.0
average = defaultdict(lambda : 0.0)
for line in iter(sys.stdin.readline, ""):
label, sent = get_label_and_sentence(line)
phi = cf.create_features(sent)
pre_label = po.predict_one(weight, phi)
if int(pre_label) != int(label):
update_weights(weight, phi, int(label))
updates += 1.0
for key, value in weight.items():
average[key] = (average[key] * (updates - 1.0) + value) / updates
# copy to weight from average
for key in weight.keys():
weight[key] = average[key]
def predict_all(model_file):
weight = defaultdict(lambda : 0.0)
# load model_file
fin = open(model_file)
for line in iter(fin.readline, ""):
parts = line.rstrip("\n").split()
value = parts.pop()
name = " ".join(parts)
weight[name] = float(value)
fin.close()
# predict all
for line in iter(sys.stdin.readline, ""):
phi = cf.create_features(line)
y = po.predict_one(weight, phi)
print y
def predict_all(model_file):
weight = defaultdict(lambda : 0.0)
# load model_file
fin = open(model_file)
for line in iter(fin.readline, ""):
parts = line.rstrip("\n").split()
value = parts.pop()
name = " ".join(parts)
weight[name] = float(value)
fin.close()
# predict all
for line in iter(sys.stdin.readline, ""):
phi = cf.create_features(line)
y = po.predict_one(weight, phi)
print int(y)
def predict_nn(network, phi):
max_layer = 1
for (name, layer, weight) in network:
max_layer = max(max_layer, layer)
# Init y
y = [{} for i in range(max_layer+1)]
y[0] = phi
# Run perceptron at each layer.
for (name, layer, weight) in network:
answer = po.predict_one(weight, y[layer-1])
y[layer][name] = answer
# The value of last layer is output of nn.
last_perceptron_name = len(network)-1
return (y[-1][last_perceptron_name], y)
def predict_nn(network, phi):
max_layer = 1
for (name, layer, weight) in network:
max_layer = max(max_layer, layer)
# Init y
y = [{} for i in range(max_layer + 1)]
y[0] = phi
# Run perceptron at each layer.
for (name, layer, weight) in network:
answer = po.predict_one(weight, y[layer - 1])
y[layer][name] = answer
# The value of last layer is output of nn.
last_perceptron_name = len(network) - 1
return (y[-1][last_perceptron_name], y)
