-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathVSBO_run.py
155 lines (144 loc) · 6.54 KB
/
VSBO_run.py
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
150
151
152
153
154
155
from VSBO_class import *
import argparse
parser = argparse.ArgumentParser('VS-BO')
parser.add_argument('--obj_func', type=str)
parser.add_argument('--method', type=str)
parser.add_argument('--momentum',type=int,default=1)
parser.add_argument('--sampling',type=str,default='CMAES_posterior')
#parser.add_argument('--num_target',type=int,default=64)
#parser.add_argument('--epochs',type=int,default=4000)
#parser.add_argument('--sgld_gamma',type=float,default=0.35)
#parser.add_argument('--folder_index',type=int)
args = parser.parse_args()
def makedirs(dirname):
if not os.path.exists(dirname):
os.makedirs(dirname)
### some parameters
N_FS = 20
acq_optim_method = 'LBFGS'
### use CMAES to sample unimportant variables
less_important_sampling = args.sampling
init_samples = 5
if args.obj_func=="Branin":
### Branin test with D=50 and d_{e}=[2,2,2]
object_dim = 50
object_bounds = torch.cat([generate_branin_bounds(2),generate_branin_bounds(2),generate_branin_bounds(object_dim-4)],dim=1)
object_func = Combine_func
function_kwargs = {'func_set':[Branin_hd,Branin_hd,Branin_hd],'var_num_set':[2,2,2],'coeff_set':[1,0.1,0.01]}
total_budget = 210
total_time = 600
total_cput = 9000
elif args.obj_func=="Hartmann6":
### Hartmann6 test with D=50 and d_{e}=[6,6,6]
object_dim = 50
object_bounds = generate_hartmann_bounds(object_dim)
object_func = Combine_func
function_kwargs = {'func_set':[Hartmann6,Hartmann6,Hartmann6],'var_num_set':[6,6,6],'coeff_set':[1,0.1,0.01]}
total_budget = 210
total_time = 400
total_cput = 7000
elif args.obj_func=="StyblinskiTang4":
### StyblinskiTang4 test with D=50 and d_{e}=[4,4,4]
object_dim = 50
object_bounds = generate_StyblinskiTang_bounds(object_dim)
object_func = Combine_func
function_kwargs = {'func_set':[StyblinskiTang4_hd,StyblinskiTang4_hd,StyblinskiTang4_hd],'var_num_set':[4,4,4],'coeff_set':[1,0.1,0.01]}
total_budget = 210
total_time = 2000
total_cput = 30000
elif args.obj_func=="rover":
from rover_test_utils import *
object_dim = 60
object_bounds = generate_rover_bounds()
object_func = rover_func
total_budget = 210
total_time = 2000
total_cput = -1
function_kwargs={}
elif args.obj_func=="mopta":
from mopta_utils import *
object_dim = 124
object_bounds = generate_mopta_bounds()
object_func = mopta_func
total_budget = 210
total_time = 3000
total_cput = -1
function_kwargs={}
output_path = "./"+args.obj_func+"/"+args.method+"/"
if args.method=="VSBO" and args.momentum==0:
output_path = "./"+args.obj_func+"/"+args.method+"_nomom/"
if args.method=="VSBO" and args.momentum==0 and args.sampling!='CMAES_posterior':
output_path = "./"+args.obj_func+"/"+args.method+"_nomom_"+args.sampling+"/"
makedirs(output_path)
if args.method =="VSBO":
#print(total_cput)
for test_id in range(1,21):
BO_instance = VSBO(N_FS,object_dim,object_func,obj_func_kwargs=function_kwargs,bounds=object_bounds)
BO_instance.data_initialize()
if(less_important_sampling=='CMAES_posterior'):
BO_instance.CMAES_initialize()
Times = []
T_process = []
F_importance_val = []
F_rank = []
F_chosen = []
t0 = time.time()
t1 = time.process_time()
iter_num = 0
#while time.time() - t0 < total_time:
while (time.time() -t0 < total_time or iter_num < total_budget or time.process_time()-t1 < total_cput):
#for one_budget in range(total_budget):
iter_num+=1
try:
### GP fitting on important variables
BO_instance.GP_fitting_active(GP_Matern)
BO_instance.BO_acq_optim_active(optim_method=acq_optim_method)
### sampling on unimportant variables
BO_instance.data_update(method=less_important_sampling,n_sampling=20)
Times.append(time.time()-t0)
T_process.append(time.process_time()-t1)
except ValueError as e:
if(e.args[0]=='Too many cov mat singular!'):
BO_instance.erase_last_instance()
iter_num-=1
continue
else:
raise ValueError(e.args[0])
if(iter_num%BO_instance.N_FS==0):
try:
BO_instance.GP_fitting(GP_Matern)
### We provide three methods for variable seletion
### KLrel: the Grad-IS method introduced in our manuscript
### ard: Automatic Relevence Determination, use the correlation length scales in the kernel function
### fANOVA: use the functional ANOVA (https://pypi.org/project/fanova/)
if args.momentum == 1:
BO_instance.variable_selection_2('KLrel')
elif args.momentum == 0:
BO_instance.variable_selection_nomom('KLrel')
#BO_instance.variable_selection('KLrel')
except ValueError as e:
if(e.args[0]=='Too many cov mat singular!'):
BO_instance.erase_last_instance()
iter_num-=1
continue
else:
raise ValueError(e.args[0])
F_importance_val.append(BO_instance.FS_important_scores)
F_rank.append(BO_instance.indices)
F_chosen.append(BO_instance.active_f_list)
if(less_important_sampling=='CMAES_posterior'):
BO_instance.CMAES_update()
print(BO_instance.active_f_list)
print(BO_instance.active_f_dims)
if(iter_num%10==0):
print(
f"Epoch {iter_num:>3} "
f"Best value: {torch.max(BO_instance.Y).item():>4.3f}"
)
np.save(output_path+"X_"+str(test_id)+".npy",BO_instance.X.numpy())
np.save(output_path+"Y_"+str(test_id)+".npy",BO_instance.Y.numpy())
np.save(output_path+"Time_"+str(test_id)+".npy",np.array(Times))
np.save(output_path+"Time_process_"+str(test_id)+".npy",np.array(T_process))
np.save(output_path+"F_importance_val_"+str(test_id)+".npy",torch.cat(F_importance_val).reshape(len(F_importance_val),object_dim).detach().numpy())
np.save(output_path+"F_rank_"+str(test_id)+".npy",torch.cat(F_rank).reshape(len(F_rank),object_dim).numpy())
np.save(output_path+"F_chosen_"+str(test_id)+".npy",torch.cat(F_chosen).reshape(len(F_chosen),object_dim).numpy())