【NetIngraph代码解读】

变量	维度	解释
adj_lists_ll	(32,2,edge_nums)	32张图，(2,edge_nums)
features_ll	(32,node_nums,13)	32张图，(node_nums,13) ：每个节点13个特征
net_labels_ll	（32,node_nums）	32张图，每张图有node_nums个节点
edgesAttr_ll	(32,44,edge_nums)	32张图，（44，edge_nums）：每个edge_nums有44个特征。

GNN/sGAT.py 是只使用节点特征
customizedGNN/wireGraph.py 用了节点特征、边特征

This code is corresponding to : It means from the shortest net length(close to zero) to the 95 percentile largest net length.The top 5% longest nets are excluded to prevent an extraordinarily large range.

y_train_95 = labels[i] > np.percentile(labels[i], 95)

output the [net_name ,net_preds] with pickle format.
Firstly, load the preds.Then, ravelse the net_list and use a list to store the net_name.

from preprocess.parse_net_new import parseRaw
nets_list = parseRaw("/media/jzt/新加卷/EDACode/NetInGraph/preprocess/dataT_folder/b11_0_mode_0.csv")
nets_names = []
for i, n in enumerate(nets_list):
    nets_names.append(n.netname)

加载pickle数据：【net_name,preds】、【net_name,labels】:

import pickle
with open("outs/b11_0_mode_0_preds.pickle","rb") as f:
    nets_names, preds = pickle.load(f)
with open("outs/b11_0_mode_0_labels.pickle","rb") as f:
    nets_names, labels = pickle.load(f)

输出pickle格式数据：【net_name,preds】、【net_name,labels】:

import pickle
with open("output/test.pickle", "wb") as f:
    pickle.dump([nets_names, preds], f)

计算preds和labels的相关度：

from scipy import stats
r = stats.pearsonr(labels, preds)[0]