from parse_data import yield_data, train, test
import numpy as np
from sklearn import preprocessing
N = 145231
if __name__ == '__main__':
X = np.empty((N, 2006))
y = np.empty(N)
for (ind, row) in enumerate(yield_data(train)):
ID, target, features = row
X[ind, :] = features
y[ind] = target
X_scaled = preprocessing.scale(X)
OUTPUT:
logarithmic loss of p given y
'''
p = max(min(p, 1. - 10e-15), 10e-15)
return -log(p) if y == 1. else -log(1. - p)
start = datetime.now()
learner = ftrl_proximal(alpha, beta, L1, L2, D, interaction)
for e in range(epoch):
loss = 0.
count = 0
for t, y, x in yield_data(train, hashing_trick=True, d=D): # data is a generator
p = learner.predict(x)
loss += logloss(p, y)
learner.update(x, p, y)
count+=1
if count%1000==0:
#print count,loss/count
print('%s\tencountered: %d\tcurrent logloss: %f' % (
datetime.now(), count, loss/count))
#if count>10000: # comment this out when you run it locally.
#break
count=0
loss=0
#import pickle