data.py with comment

GCN-PyTorch-master/data.py

@@ -11,6 +11,7 @@ def parse_index_file(filename):
     Parse index file.
     """
     index = []
+    # 解析index文件的，传进来的是ind.{}.test.index文件
     for line in open(filename):
         index.append(int(line.strip()))
     return index
@@ -20,6 +21,7 @@ def sample_mask(idx, l):
     """
     Create mask.
     """
+    # 生成一个行向量，其中，idx对应的部分为True 其他部分False
     mask = np.zeros(l)
     mask[idx] = 1
     return np.array(mask, dtype=np.bool)
@@ -29,17 +31,27 @@ def load_data(dataset_str):
     """
     Loads input data from gcn/data directory
 
+    https://blog.csdn.net/yyl424525/article/details/100831452（前六个讲的很好）
+    https://blog.csdn.net/chl183/article/details/107446836（ind.dataset_str.graph用的对象，他这个反应的是节点之间的关系：边信息）
+    那个defaultdict：dict_items([('yellow', [1, 3]), ('blue', [2, 4])]
     ind.dataset_str.x => the feature vectors of the training instances as scipy.sparse.csr.csr_matrix object;
+    scipy.sparse.csr.csr_matrix object：稀疏矩阵：（location） value，x是训练集特征
     ind.dataset_str.tx => the feature vectors of the test instances as scipy.sparse.csr.csr_matrix object;
+    测试集特征
     ind.dataset_str.allx => the feature vectors of both labeled and unlabeled training instances
         (a superset of ind.dataset_str.x) as scipy.sparse.csr.csr_matrix object;
+    all指的是剩下的 的特征，因为是半监督的分类，因而
     ind.dataset_str.y => the one-hot labels of the labeled training instances as numpy.ndarray object;
+    one-hot编码，训练集的分类
     ind.dataset_str.ty => the one-hot labels of the test instances as numpy.ndarray object;
+    one-hot编码，测试集的分类
     ind.dataset_str.ally => the labels for instances in ind.dataset_str.allx as numpy.ndarray object;
+    全部的费雷
     ind.dataset_str.graph => a dict in the format {index: [index_of_neighbor_nodes]} as collections.defaultdict
         object;
+    这个东西就是图的链表表示在Python上边的东西
     ind.dataset_str.test.index => the indices of test instances in graph, for the inductive setting as list object.
-
+    测试机的编号
     All objects above must be saved using python pickle module.
 
     :param dataset_str: Dataset name
@@ -55,8 +67,12 @@ def load_data(dataset_str):
                 objects.append(pkl.load(f))
 
     x, y, tx, ty, allx, ally, graph = tuple(objects)
-    test_idx_reorder = parse_index_file("data/ind.{}.test.index".format(dataset_str))
-    test_idx_range = np.sort(test_idx_reorder)
+    # 把前几个数据读进来
+    test_idx_reorder = parse_index_file("data/ind.{}.test.index".format(dataset_str))       # reorder 乱序的
+    test_idx_range = np.sort(test_idx_reorder)                                              # range 有序的
+    # 把test的数据读进来
+    # 上边几个完成了数据从文件中读进来
+
 
     if dataset_str == 'citeseer':
         # Fix citeseer dataset (there are some isolated nodes in the graph)
@@ -69,20 +85,22 @@ def load_data(dataset_str):
         ty_extended[test_idx_range-min(test_idx_range), :] = ty
         ty = ty_extended
 
-    features = sp.vstack((allx, tx)).tolil()
-    features[test_idx_reorder, :] = features[test_idx_range, :]
-    adj = nx.adjacency_matrix(nx.from_dict_of_lists(graph))
+    features = sp.vstack((allx, tx)).tolil()                        # 先合并起来，然后再将他转换为系数矩阵的矩阵表示形式
+    features[test_idx_reorder, :] = features[test_idx_range, :]     # 将特征和编号对应起来
+    # https://zhuanlan.zhihu.com/p/356227823
+    adj = nx.adjacency_matrix(nx.from_dict_of_lists(graph))         # 构建联通矩阵
 
-    labels = np.vstack((ally, ty))
-    labels[test_idx_reorder, :] = labels[test_idx_range, :]
+    labels = np.vstack((ally, ty))                                  # ty和ally共同构成的label的组合体
+    labels[test_idx_reorder, :] = labels[test_idx_range, :]         # 对应label和序列的index
 
-    idx_test = test_idx_range.tolist()
-    idx_train = range(len(y))
+    idx_test = test_idx_range.tolist()                              # 记录test（包括tx和allx）的index
+    idx_train = range(len(y))                                       # 记录train的index
     idx_val = range(len(y), len(y)+500)
 
     train_mask = sample_mask(idx_train, labels.shape[0])
     val_mask = sample_mask(idx_val, labels.shape[0])
     test_mask = sample_mask(idx_test, labels.shape[0])
+    # 相当于是给了一个索引，用mask去筛选数据
 
     y_train = np.zeros(labels.shape)
     y_val = np.zeros(labels.shape)
@@ -90,7 +108,9 @@ def load_data(dataset_str):
     y_train[train_mask, :] = labels[train_mask, :]
     y_val[val_mask, :] = labels[val_mask, :]
     y_test[test_mask, :] = labels[test_mask, :]
+    # 筛选出来的数据去进行对应，放进对应名字里边去
 
+    # https://blog.csdn.net/qq_37995260/article/details/100146401 几个返回值分别的作用
     return adj, features, y_train, y_val, y_test, train_mask, val_mask, test_mask
 
 
@@ -128,7 +148,9 @@ def preprocess_features(features):
 
 
 def normalize_adj(adj):
+    # 这个东西的返回值就是那个系数D^-0.5AD^0.5
     """Symmetrically normalize adjacency matrix."""
+    # scipy.sparse as sp
     adj = sp.coo_matrix(adj)
     rowsum = np.array(adj.sum(1)) # D
     d_inv_sqrt = np.power(rowsum, -0.5).flatten() # D^-0.5
@@ -138,6 +160,7 @@ def normalize_adj(adj):
 
 
 def preprocess_adj(adj):
+    #这个是实现了那个A+1的那一步
     """Preprocessing of adjacency matrix for simple GCN model and conversion to tuple representation."""
     adj_normalized = normalize_adj(adj + sp.eye(adj.shape[0]))
     return sparse_to_tuple(adj_normalized)
@@ -147,12 +170,16 @@ def preprocess_adj(adj):
 
 
 def chebyshev_polynomials(adj, k):
+    # 切比雪夫是用作卷积核的，这个我之前一直以为是推前向传播的，有点干蒙了，离谱，本质上就是一个转换成前向传播函数的系数求解
+    # k为切比雪夫的阶数
+    # 这玩意就是没用到啊。。。。。。，在文件里都没出现，无语。他只用了两个preprocess函数
+    # 看明白了，教学用的
     """
     Calculate Chebyshev polynomials up to order k. Return a list of sparse matrices (tuple representation).
     """
     print("Calculating Chebyshev polynomials up to order {}...".format(k))
 
-    adj_normalized = normalize_adj(adj)
+    adj_normalized = normalize_adj(adj)                     # 返回那个正向传播的那个的系数矩阵D^-0.5AD^0.51
     laplacian = sp.eye(adj.shape[0]) - adj_normalized
     largest_eigval, _ = eigsh(laplacian, 1, which='LM')
     scaled_laplacian = (2. / largest_eigval[0]) * laplacian - sp.eye(adj.shape[0])