optimize/sparse_owlqn.py at master · summersunshine1/optimize · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
import numpy as np
from getPath import *
pardir = getparentdir()
from commonLib import *
from adadelta import *
from sklearn import metrics
from commonmethod import *

train_path = pardir+'/data/train.ffm'
test_path = pardir+'/data/test.ffm'
auc_path = pardir+'/res/auc_owl_qn_10'
e = 1e-6
lamda = 0.0001

def reshape_arr(arr):
    return np.squeeze(arr).A1

def compfunc(v,lamda):
    if v+lamda<0:
        return [v+lamda]
    elif v-lamda>0:
        return [v-lamda]
    return 0


def get_pesudo_gradient(realg,w):
    # shape = np.shape(realg)
    g = np.array([[0.0]]*len(w))
    indexs = np.array(list(range(len(w))))
    # lesszero_index = indexs[reshape_arr(w<0)]
    # morezero_index = indexs[reshape_arr(w>0)]
    zero_index = indexs[reshape_arr(w==0)]
    less = reshape_arr((w<0))
    g[less] = realg[less]-lamda
    more = reshape_arr((w>0))
    g[more] = realg[more]+lamda

    leftindex = zero_index[reshape_arr((realg[zero_index]<-lamda))]
    rightindex = zero_index[reshape_arr((realg[zero_index]>lamda))]
    g[leftindex] = realg[leftindex]+lamda
    g[rightindex] = realg[rightindex]-lamda
    g = np.matrix(g)
    return g

def get_orthant(w,psudo_g):
    orthant = w.copy()
    nozeroindex = (w!=0)
    zeroindex = (w==0)
    nozerosign = np.sign(orthant[nozeroindex])
    if len(nozerosign)>0:
        orthant[nozeroindex]=nozerosign.A1
    zerosign = np.sign(-psudo_g[zeroindex])
    if len(zerosign)>0:
        orthant[zeroindex]=zerosign.A1
    return orthant

def fix_sign(g,sign):
    res = np.multiply(g,sign)
    g[res<=0]=0.0
    return g

def online_owl_qn(w,vecfeatures,labels):
    epsilo = 1e-4
    n = np.shape(w)[0]
    c = 1
    k = 0
    m = 10
    s = []
    y = []

    batch_size = 100
    local_lamda = 0.1
    minimum = 1e-10

    t0 =np.power(10,4)
    ada = Adam(n,alpha=0.01)

    iter = 0
    lr = 0.01

    while k<100:
        j = 0
        while j+batch_size<=len(labels):
            iter+=1
            # print("current iter:" +str(iter))
            realg = compute_regular_gradients(vecfeatures[j:j+batch_size],labels[j:j+batch_size],w)
            pg = get_pesudo_gradient(realg,w)
            # pg = realg
            if j==0 and k==0:
                d = -pg*minimum
            d = fix_sign(d,-pg)
            templr = ada.getmaxgrad(d,iter)
            orth = get_orthant(w,pg)

            new_w = w+templr
            new_w = fix_sign(new_w,orth)
            sk = (new_w - w)

            newg = compute_regular_gradients(vecfeatures[j:j+batch_size],labels[j:j+batch_size,:],new_w)
            yk = newg-realg+local_lamda*sk
            w = new_w.copy()
            if len(s)>m:
                s.pop(0)
                y.pop(0)
            s.append(sk)
            y.append(yk)

            begin = time.time()
            d = lbfgs_two_recursion(s,y,pg,d,c=1)
            end = time.time()
            print_consume_time(begin,end,"two recursion",isprint=0)
            # d = fix_sign(d,-pg)

            j+=batch_size
            if iter%10 == 0:
                test(w,test_features,test_labels,auc_path,isl1=1)
        k+=1
    return w

test_features = 0
test_labels = 0
train_features = 0
train_labels = 0

def train():
    if os.path.exists(auc_path):
        os.remove(auc_path)
    global train_features,train_labels
    train_features,train_labels,w = initdata(test_path)

    global test_features,test_labels
    test_features,test_labels,_ = initdata(test_path)
    maxiter = 10
    w = online_owl_qn(w,train_features,train_labels)

if __name__=="__main__":
    train()