RBM/RBMI.py at main · RaymndH/RBM · 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
from sklearn.neural_network import BernoulliRBM
from sklearn.pipeline import Pipeline
from sklearn.linear_model import LogisticRegression
from sklearn import datasets
import numpy as np
from scipy.special import expit

def get_en(v):
    return .5*(np.dot(v.T,np.dot(wmat,v)))

def Is_en(v):
    return .5*(np.dot(v,np.dot(wmat,v)))

def RBM_en(v,h):
    return np.dot(v.T,np.dot(W,h))

def neuron(w):
    p=np.dot(W,w)
    logp=np.sum(np.log(1+np.exp(p)))
    return p, logp

def gibbs(w, max_prob, best_state):
    p,logp=neuron(w)

    if(logp>max_prob):
        max_prob=logp
        best_state=w
    #print("logp",logp)
    expit(p, out=p)
    return 2*(np.random.uniform(size=w.shape) < p)-1, best_state, max_prob


# Load your dataset
fnames = [  "wishart_planting_N_16_alpha_0.19/wishart_planting_N_16_alpha_0.19_inst_1.txt",
            "wishart_planting_N_16_alpha_0.50/wishart_planting_N_16_alpha_0.50_inst_1.txt",
            "wishart_planting_N_16_alpha_0.75/wishart_planting_N_16_alpha_0.75_inst_1.txt",
            "wishart_planting_N_16_alpha_0.88/wishart_planting_N_16_alpha_0.88_inst_1.txt",
            "wishart_planting_N_16_alpha_1.12/wishart_planting_N_16_alpha_1.12_inst_1.txt"]
grounds = [ -1.2655691,
            -3.9861914,
            -6.987742,
            -6.2008785,
            -8.9782550]


iter = 4

ground = grounds[iter]
fname = fnames[iter]
digits = np.loadtxt(fname)
N=16
wmat=np.zeros((16,16))
for line in digits:
    wmat[int(line[0]), int(line[1])] = line[2]
wmat=wmat+wmat.T

W = -5*wmat

dotones1 = np.dot(W,np.ones(N))


alpha = 1.1

C = -alpha * np.abs(dotones1)
#C = -.25*np.ones(N)
#C = -.5*np.ones(N)

for i in range(N):
    W[i,i]=-C[i]


# Create an RBM model

plant=np.array([-1, -1, -1, -1, 1, -1, 1, -1, 1, 1, 1, 1, -1, 1, -1, 1])
print("plant is \n",plant)
successes=0
N_trials=19
success_states=[]
print("beginning trials:")
for trial in range(N_trials):
    if(1):
        v=np.random.choice([-1,1],[16])
        h=np.random.choice([-1,1],[16])
        #
    else:
        v=np.random.choice([-1,1],[16,1])
        v=np.ones_like(v)
        h=v
    #h=v

    #print("initial state of v is", v)
    #print("initial state of h is", h)
    #print()
    min_found=100
    max_prob=-100
    best_state=np.zeros_like(v)
    bester=np.zeros_like(v)
    for i in range(10000):

        #print("energy is", RBM_en(v,h))
        #print("other energy is", get_en(v))
        #print("ising energy is", Is_en(v))
        if(Is_en(v)<min_found):
            min_found=Is_en(v)
            bester=v
            #print(v.T.astype(int))
        #forward gibbs sample
        h,best_state,max_prob=gibbs(v,max_prob, best_state)
        #print(h.T.astype(int))
        v,best_state,max_prob=gibbs(h,max_prob, best_state)

        #backwards gibbs sample


    #print("max p state", best_state)
    #print("diff",best_state-plant)
    #print("bestest",bester)
    #print("")
    print("trial",trial,"minimum energy found: ",min_found)
    print("Ising energy of highest probability state : ", Is_en(best_state))
    print()
    if (Is_en(best_state) < ground + 1e-3):
        successes+=1
        success_states.append(best_state)

print(successes/N_trials, " success rate")
print()