Collaborative-Deep-Learning-for-Recommender-Systems/mf_auto_v2.py at master · xiaoouzhang/Collaborative-Deep-Learning-for-Recommender-Systems · 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
139
140
141
142
143
144
145
146
147
148
149
import h5py
import numpy as np
import random
import auto_fun as auto
import time

INPUT_LAYER = 314
HIDDEN_UNIT1 = 200
HIDDEN_UNIT2 = 40
LEARNING_RATE = 0.001/20
#number of epoch for runing autoencoder in each interation
EPOCH_enc = 30
#number of mf updates in each iteration
EPOCH_mf=10
#randomSeed = np.random.RandomState(42)
mu, sigma = 0, 0.1
l=HIDDEN_UNIT2
alpha=40
l2_u=100.0
l2_v=600.0
batch=500
ratio_l=1000.0
ratio_u=2.0
l1=0.01

def main(denoise = True):
    #allMatrix, xtrain, xval, xtest,lenList,accList = getData()
    diction = [('ind_empleado', 5), ('pais_residencia', 24), ('sexo', 3), ('ind_nuevo', 2), ('indrel', 2), ('indrel_1mes', 4), ('tiprel_1mes', 4), ('indresi', 2), ('indext', 2), ('conyuemp', 3), ('canal_entrada', 158), ('indfall', 2), ('cod_prov', 53), ('ind_actividad_cliente', 2), ('segmento', 4), ('antiguedad_binned', 10), ('age_binned', 24), ('renta_binned', 10)]
    lenList = []
    for tuppl in diction:
        val = tuppl[1]
        lenList.append(val)
    accList = []
    for i in range(len(lenList)):
        if i ==0:
            accList.append(lenList[i])
        else:
            accList.append(accList[i-1]+lenList[i])
    #read user infor
    with h5py.File('user_infor.h5', 'r') as hf:
        xtrain = hf['infor'][:]
    #with h5py.File('user_infor_new.h5', 'r') as hf:
    #    x_new = hf['infor'][:]
    #read rating matrix
    with h5py.File('rating_tr_numpy.h5', 'r') as hf:
        rating_mat = hf['rating'][:]

    W1,W2,b1,b2,c1,c2 = auto.initialization(INPUT_LAYER,[HIDDEN_UNIT1,HIDDEN_UNIT2],mu,sigma)
    #define user and item matrices
    u=np.random.rand(rating_mat.shape[0],l)
    v=np.random.rand(rating_mat.shape[1],l)
    '''
    with h5py.File('u_120_40bi.h5', 'r') as hf:
        u = hf['u'][:]
    with h5py.File('v_120_40bi.h5', 'r') as hf:
        v = hf['v'][:]
    with h5py.File('W1_120_40bi.h5', 'r') as hf:
        W1 = hf['W1'][:]
    with h5py.File('b1_120_40bi.h5', 'r') as hf:
        b1 = hf['b1'][:]
    with h5py.File('c1_120_40bi.h5', 'r') as hf:
        c1 = hf['c1'][:]
    with h5py.File('W2_120_40bi.h5', 'r') as hf:
        W2 = hf['W2'][:]
    with h5py.File('b2_120_40bi.h5', 'r') as hf:
        b2 = hf['b2'][:]
    with h5py.File('c2_120_40bi.h5', 'r') as hf:
        c2 = hf['c2'][:]
    '''
    #define preference and confidence matrices
    p=np.zeros(rating_mat.shape)
    p[rating_mat>0]=1
    c=np.zeros(rating_mat.shape)
    c=1+alpha*rating_mat

    iteration=30

    print('start')
    for iterate in range(iteration):
        for iter_mf in np.arange(EPOCH_enc):

            #update v
            start=time.time()

            u2=u.T.dot(u)
            for j in range(rating_mat.shape[1]):
                #print(j)
                nonzero_list=np.nonzero(p[:,j])[0]
                c_diag_nonzero=np.diag(c[nonzero_list,j]-1)
                u_nonzero=u[nonzero_list,:]

                temp_v=np.dot(c[nonzero_list,j],u_nonzero)
                v[j,:]=np.dot(temp_v,np.linalg.pinv(l2_v*np.identity(l)+np.dot(np.dot(u_nonzero.T,c_diag_nonzero),u_nonzero)+u2))
            print('v complete')

            end=time.time()
            print(end-start)

            #update u
            start=time.time()
            v2=v.T.dot(v)
            for i in range(rating_mat.shape[0]):

                nonzero_list=np.nonzero(p[i,:])[0]
                v_nonzero=v[nonzero_list,:]
                c_diag_nonzero=np.diag(c[i,nonzero_list]-1)

                temp_u=np.dot(c[i,nonzero_list],v_nonzero)
                u[i,:]=np.dot(temp_u,np.linalg.pinv(l2_u*np.identity(l)+np.dot(np.dot(v_nonzero.T,c_diag_nonzero),v_nonzero)+v2))
            print('u complete')
            end=time.time()
            print(end-start)

            print(np.linalg.norm(p-np.dot(u,v.T)))

        W1,W2,b1,b2,c1,c2 = auto.autoEncoder(ratio_l,ratio_u,batch,W1,W2,xtrain,u,b1,b2,c1,c2,accList,EPOCH_enc,LEARNING_RATE,l1,denoise = True)
        hidden =  auto.getoutPut(W1,W2,b1,b2,xtrain,accList)
        print(np.sum(np.square(u)))
        print(np.sum(np.square(u-hidden)))
        u=hidden
        print(np.linalg.norm(p-np.dot(u,v.T)))

    with h5py.File('u_200_40bi.h5', 'w') as hf:
        hf.create_dataset("u",  data=u)
    with h5py.File('v_200_40bi.h5', 'w') as hf:
        hf.create_dataset("v",  data=v)
    with h5py.File('W1_200_40bi.h5', 'w') as hf:
        hf.create_dataset("W1",  data=W1)
    with h5py.File('b1_200_40bi.h5', 'w') as hf:
        hf.create_dataset("b1",  data=b1)
    with h5py.File('c1_200_40bi.h5', 'w') as hf:
        hf.create_dataset("c1",  data=c1)
    with h5py.File('W2_200_40bi.h5', 'w') as hf:
        hf.create_dataset("W2",  data=W2)
    with h5py.File('b2_200_40bi.h5', 'w') as hf:
        hf.create_dataset("b2",  data=b2)
    with h5py.File('c2_200_40bi.h5', 'w') as hf:
        hf.create_dataset("c2",  data=c2)

    #making_graph(trainLoss, valiLoss, testLoss)
    #
    # hidden = pd.DataFrame(hidden.T)
    # prediction = pd.DataFrame(prediction.T)
    # hidden.to_csv('hidden.csv')
    # prediction.to_csv('prediction.csv')

    return hidden

main(denoise = True)