GraphBerry / DiG-In-GNN

The official implementation of DiG-In-GNN (AAAI 2024)
17 stars 3 forks source link

你好,想问一下,这里不是利用了原始标签进行训练的吗,为什么还是无监督学习呢。论文中提出是划分Vb,所以没有太看明白,想请教一下 #1

Closed angle2001 closed 1 month ago

angle2001 commented 3 months ago
def get_contrast_sample(self, gen_feats, raw_feats, env_feats, env_raw_feats, labels):

classify_label_1 = [] classify_label_0 = [] for i in range(len(labels)): if labels[i] == torch.tensor(1): classify_label_1.append(i) else: classify_label_0.append(i)

XZJIsme commented 2 months ago

你好,本文不是无监督学习,对比学习的时候使用了训练集的标签; $\mathcal{V}_b$ 是一个 batch,本文是 mini-batch,请问具体的疑惑是?