运行DGBO_GC-batch

[ghost]

崔老师您好，我在学习您的DGBO_GC方法，我采用的overfit_type是默认的opt。目前程序运行初始化参数，代码如下：
def initialization(data,n,info):
print ("-----------initialization-----------")
xlist=np.linspace(0,len(data.candidates)-1,len(data.candidates))
xlist = xlist.tolist()
xlistinit=[]
if len(data.candidates)%n==0:
delta=len(data.candidates)/n
else:
delta=len(data.candidates)/n+1
for i in range(n):
xlistinit.append(random.sample(xlist[i*int(delta):min((i+1)*int(delta),len(data.candidates))],1)[0])
xlist=xlistinit
xlist=np.array(xlist,dtype=np.integer)
for i in range(n):
print ("initialization [%s] ..."%(i))
y=evaluate(info,data.candidates[xlist[i]])
info.add(xlist[i],y)
if info.overfit_type=="opt":
print ("init the setting of params on overfitting...")
bo = BayesianOptimization(initsettingCV,{'dropout_': (0,1), 'weight_decay_': (1e-5,1e-1)},random_state=rng)
bo.maximize(init_points=21, n_iter=30, acq='ei')
print (bo.res['max'])
在上述代码中，运行到 bo = BayesianOptimization(initsettingCV,{'dropout_': (0,1), 'weight_decay_': (1e-5,1e-1)},random_state=rng)后进入了5折交叉验证，但是一直循环交叉验证的过程，结果如下：
| 2 | -1.044 | 0.9985 | 0.01969 |
[cv 1] [ 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19] [0 1 2 3]
add a model
[cv 2] [ 0 1 2 3 8 9 10 11 12 13 14 15 16 17 18 19] [4 5 6 7]
add a model
[cv 3] [ 0 1 2 3 4 5 6 7 12 13 14 15 16 17 18 19] [ 8 9 10 11]
add a model
[cv 4] [ 0 1 2 3 4 5 6 7 8 9 10 11 16 17 18 19] [12 13 14 15]
add a model
[cv 5] [ 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15] [16 17 18 19]
add a model
transition_matrix: (16, 257)
train: features_comb shape is : (257, 68)
train: constructing adj...
Build DGCN_Comb2...
Architecture: { gcn[1 2 3 4 5]->concat(0.8)->dnn[1 2 3 4 5] }
heur: dropout=0.16990029785933825, weight_decay=0.047577751206738225
Start to train network.
Epoch: 0001 train_loss= 45.74941 time= 1.25621s total_time= 1.25621s
Epoch: 0101 train_loss= 33.50009 time= 0.00600s total_time= 1.88035s
Train net Finished!
acc= 1.3567712
transition_matrix: (16, 257)
train: features_comb shape is : (257, 68)
train: constructing adj...
Build DGCN_Comb2...
Architecture: { gcn[1 2 3 4 5]->concat(0.8)->dnn[1 2 3 4 5] }
heur: dropout=0.16990029785933825, weight_decay=0.047577751206738225
Start to train network.
Epoch: 0001 train_loss= 53.56653 time= 1.30929s total_time= 1.30929s
Epoch: 0101 train_loss= 33.76947 time= 0.00700s total_time= 1.95644s
Train net Finished!
acc= 1.1174575
transition_matrix: (16, 257)
train: features_comb shape is : (257, 68)
train: constructing adj...
Build DGCN_Comb2...
Architecture: { gcn[1 2 3 4 5]->concat(0.8)->dnn[1 2 3 4 5] }
heur: dropout=0.16990029785933825, weight_decay=0.047577751206738225
Start to train network.
Epoch: 0001 train_loss= 51.28548 time= 1.23025s total_time= 1.23025s
Epoch: 0101 train_loss= 33.82351 time= 0.00700s total_time= 1.86339s
Train net Finished!
acc= 0.5193358
transition_matrix: (16, 257)
train: features_comb shape is : (257, 68)
train: constructing adj...
Build DGCN_Comb2...
Architecture: { gcn[1 2 3 4 5]->concat(0.8)->dnn[1 2 3 4 5] }
heur: dropout=0.16990029785933825, weight_decay=0.047577751206738225
Start to train network.
Epoch: 0001 train_loss= 45.57104 time= 1.31298s total_time= 1.31298s
Epoch: 0101 train_loss= 32.98197 time= 0.00700s total_time= 1.97312s
Train net Finished!
acc= 1.7546887
transition_matrix: (16, 257)
train: features_comb shape is : (257, 68)
train: constructing adj...
Build DGCN_Comb2...
Architecture: { gcn[1 2 3 4 5]->concat(0.8)->dnn[1 2 3 4 5] }
heur: dropout=0.16990029785933825, weight_decay=0.047577751206738225
Start to train network.
Epoch: 0001 train_loss= 47.50560 time= 1.25738s total_time= 1.25738s
Epoch: 0101 train_loss= 33.11481 time= 0.00600s total_time= 1.89652s
Train net Finished!
acc= 1.9915786
上述过程一直循环，无法运行到 bo.maximize(init_points=21, n_iter=30, acq='ei')这一步。不知道问题出现在哪里。
另外发现当不debug，直接运行整个项目，当迭代次数为51之后会跳到bo.maximize(init_points=21, n_iter=30, acq='ei')这一步，但是会报错。错误如下：Traceback (most recent call last):
File "E:/model/Scalable-and-Parallel-DGBO-master/Scalable-and-Parallel-DGBO-master/DGBO_GC-batch/DGBO.py", line 338, in
BO_process(data,params,info)
File "E:/model/Scalable-and-Parallel-DGBO-master/Scalable-and-Parallel-DGBO-master/DGBO_GC-batch/DGBO.py", line 210, in BO_process
initialization(data,params.initn,info)
File "E:/model/Scalable-and-Parallel-DGBO-master/Scalable-and-Parallel-DGBO-master/DGBO_GC-batch/DGBO.py", line 135, in initialization
print (bo.res['max'])
TypeError: list indices must be integers or slices, not str。
这个问题困扰我很久了，一直不知道原因，希望您百忙之中可以帮忙给解答一下。我将感激不尽，谢谢您~

[csjtx1021]

Hi，

Can you confirm that the installed bayesian-optimization package is version 1.0.0?

Best