def predict_all(model_file, input_file):
w = dict()
for line in open(model_file):
name, value = line.strip('\n').split('\t')
w[name] = float(value)
with open('my_answer.txt', 'w') as fp:
for x in open(input_file):
phi = create_features(x)
y_d = predict_one(w, phi)
print(y_d, file=fp)
def predict_all(model_file, input_file):
w = dict()
for line in open(model_file):
name, value = line.strip('\n').split('\t')
w[name] = float(value)
with open('my_answer.txt', 'w') as fp:
for x in open(input_file):
phi = create_features(x)
y_d = predict_one(w, phi)
print(y_d, file=fp)
def train_nn(train_file, model_file, num_nodes, num_layers, num_iter):
"""Train a neural network"""
import random
import pickle
import sys
NUM_NODES = num_nodes
NUM_LAYERS = num_layers
NUM_ITER = num_iter
# Too large learning rate gives bizarre results
#LEARNING_RATE = 0.5
LEARNING_RATE = 0.1
# See page 10
network = []
for i in range(NUM_LAYERS):
for j in range(NUM_NODES):
weight = {}
node = (i + 1, weight) # note that + 1
network.append(node)
# the last node
network.append((NUM_LAYERS + 1, {}))
for i in range(NUM_ITER):
print >>sys.stderr, "Iteration #%s..." % (i)
for line in open(train_file, 'r'):
line = line.rstrip()
(y, x) = line.split("\t")
y = int(y)
phi = create_features(x)
update_nn(network, phi, y, LEARNING_RATE)
# debugging
for line in open(train_file, 'r'):
line = line.rstrip()
(y, x) = line.split("\t")
y = int(y)
phi = create_features(x)
y_ = predict_nn(network, phi)
print y, y_[-1].values()[0]
with open(model_file, 'w') as f:
pickle.dump(network, f)
def predict_all(model_file,input_file):
with open(model_file,'r') as m_f,open(input_file,'r') as i_f:
w = defaultdict(lambda :0)
for line in m_f:
word,value = line.split('\t')
w[word] = float(value)
for x in i_f:
phi = defaultdict(lambda :0)
for k,v in create_features(x).items():
phi[k] = float(v)
y_ = predict_one(w,phi)
yield('{}\t{}'.format(y_,x))
def test_nn(test_file, model_file):
"""Test using a neural network"""
import pickle
with open(model_file, 'r') as f:
network = pickle.load(f)
for line in open(test_file, 'r'):
line = line.rstrip()
phi = create_features(line)
y = predict_nn(network, phi)
#print y[-1].values()[0]
if y[-1].values()[0] >= 0:
print 1
else:
print -1
def test_nn(test_file, model_file):
"""Test using a neural network"""
import pickle
with open(model_file, "r") as f:
network = pickle.load(f)
for line in open(test_file, "r"):
line = line.rstrip()
phi = create_features(line)
y = predict_nn(network, phi)
# print y[-1].values()[0]
if y[-1].values()[0] >= 0:
print 1
else:
print -1
def predict_all(model_file, input_file):
"""Reads model file and predict instance one by one"""
# Reading a model file
w = {}
for line in open(model_file):
line = line.strip()
(name, value) = line.split("\t")
value = float(value)
w[name] = value
# Evaluation and print results
for line in open(input_file):
line = line.strip()
phi = create_features(line)
y_ = predict_one(w, phi)
print y_
def predict_all(model_file, input_file):
"""Reads model file and predict instance one by one"""
# Reading a model file
w = {}
for line in open(model_file):
line = line.strip()
(name, value) = line.split("\t")
value = float(value)
w[name] = value
# Evaluation and print results
for line in open(input_file):
line = line.strip()
phi = create_features(line)
y_ = predict_one(w, phi)
print y_
def pred_all(model_file, input_file=inputfile):
ansfile = model_file + '_ans'
phi = defaultdict(lambda : 0.)
with open(model_file) as modelf:
w = defaultdict(lambda: 0.)
for model_line in modelf:
name, weight = model_line.split('\t')
# print("{}\t{}".format(name,weight))
w[name] = float(weight)
with open(input_file) as inputf, open(ansfile, 'w') as ansout:
for in_line in inputf:
l = in_line
# print(l)
phi = create_features(l.lower())
y_pred = int(predict_one(w, phi))
print('{}\t{}'.format(y_pred, l), end='', file=ansout)
import subprocess
subprocess.call('../../script/grade-prediction.py ../../data/titles-en-test.labeled ' + ansfile, shell=True)
from train_svm import sign, create_features, product
from collections import defaultdict
W = defaultdict(float)
for line in open('train_result.txt', 'r'):
key, value = line.split('\t')
W[key] = float(value)
for x in open('../../data/titles-en-test.word', 'r'):
phi = create_features(x)
y_prime = product(W, phi)
print(sign(y_prime))