confidence weighted learning

这个算法原作者意思是在NLP中面对高维向量和数据稀疏时效果会不错，算法保持了当前预测向量w的均值和方差，并做优化。代码实现如下：

#cw learning algorithm
def get_phi():
    '''confidence parameter phi'''
    import numpy as np
    from scipy import stats
    eta = float(raw_input("please input confidence parameter between (0 1)\n"))
    X = stats.norm(0.0,1.0)
    phi = X.ppf(eta)
    return phi

def get_data_size(datafile):
    f = open(datafile)
    size = f.readline().strip().split()
    f.close()
    return len(size)

def dot(X,Y):
    sum = 0
    for (x,y) in zip(X,Y):
        sum += x*y
    return sum
def mul(X,Y):
    Z = []
    for (x,y) in zip(X,Y):
        Z.append(x*y)
    return Z
def plus(X,Y):
    Z = []
    for (x,y) in zip(X,Y):
        Z.append(x+y)
    return Z

def opp(X):
    Y = []
    for x in X:
        Y.append(1/x)
    return Y
def init_u_E(datafile):
    #initialize
    #inp