GCN代码分析
1 代码结构
.
├── data // 图数据
├── inits // 初始化的一些公用函数
├── layers // GCN层的定义
├── metrics // 评测指标的计算
├── models // 模型结构定义
├── train // 训练
└── utils // 工具函数的定义
utils.py
def parse_index_file(filename) # 处理index文件并返回index矩阵
def sample_mask(idx, l) #创建 mask 并返回mask矩阵
def load_data(dataset_str) # 读取数据
-
从gcn/data文件夹下读取数据，文件包括有：
-
ind.dataset_str.x => 训练实例的特征向量，如scipy.sparse.csr.csr_matrix类的实例
-
ind.dataset_str.tx => 测试实例的特征向量，如scipy.sparse.csr.csr_matrix类的实例
-
ind.dataset_str.allx => 有标签的+无无标签训练实例的特征向量，是ind.dataset_str.x的超集
-
ind.dataset_str.y => 训练实例的标签，独热编码，numpy.ndarray类的实例
-
ind.dataset_str.ty => 测试实例的标签，独热编码，numpy.ndarray类的实例
-
ind.dataset_str.ally => 有标签的+无无标签训练实例的标签，独热编码，numpy.ndarray类的实例
-
ind.dataset_str.graph => 图数据，collections.defaultdict类的实例，格式为 {index：[index_of_neighbor_nodes]}
-
ind.dataset_str.test.index => 测试实例的id
上述文件必须都用python的pickle模块存储
- 返回： adj, features, y_train, y_val, y_test, train_mask, val_mask, test_mask
def sparse_to_tuple(sparse_mx) # 将矩阵转换成tuple格式并返回
def preprocess_features(features) # 处理特征:将特征进行归一化并返回tuple (coords, values, shape)
def normalize_adj(adj) # 图归一化并返回
def preprocess_adj(adj) # 处理得到GCN中的归一化矩阵并返回
def construct_feed_dict(features, support, labels, labels_mask, placeholders) # 构建输入字典并返回
def chebyshev_polynomials(adj, k) # 切比雪夫多项式近似:计算K阶的切比雪夫近似矩阵
def chebyshev_polynomials(adj, k): """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) # D^{-1/2}AD^{1/2} laplacian = sp.eye(adj.shape[0]) - adj_normalized # L = I_N - D^{-1/2}AD^{1/2} largest_eigval, _ = eigsh(laplacian, 1, which='LM') # \lambda_{max} scaled_laplacian = (2. / largest_eigval[0]) * laplacian - sp.eye(adj.shape[0]) # 2/\lambda_{max}L-I_N # 将切比雪夫多项式的 T_0(x) = 1和 T_1(x) = x 项加入到t_k中 t_k = list() t_k.append(sp.eye(adj.shape[0])) t_k.append(scaled_laplacian) # 依据公式 T_n(x) = 2xT_n(x) - T_{n-1}(x) 构造递归程序，计算T_2 -> T_k项目 def chebyshev_recurrence(t_k_minus_one, t_k_minus_two, scaled_lap): s_lap = sp.csr_matrix(scaled_lap, copy=True) return 2 * s_lap.dot(t_k_minus_one) - t_k_minus_two for i in range(2, k+1): t_k.append(chebyshev_recurrence(t_k[-1], t_k[-2], scaled_laplacian)) return sparse_to_tuple(t_k) layers.py
-
各层定义的方式与keras类似
-
定义基类 Layer
属性：name (String) => 定义了变量范围；logging (Boolean) => 打开或关闭TensorFlow直方图日志记录
方法：
__init__()(初始化)，_call()(定义计算)，__call__()(调用_call()函数)，_log_vars() -
定义Dense Layer类，继承自Layer类
-
定义GraphConvolution类，继承自Layer类。重点来看一下这个类的实现。
class GraphConvolution(Layer): """Graph convolution layer.""" def __init__(self, input_dim, output_dim, placeholders, dropout=0., sparse_inputs=False, act=tf.nn.relu, bias=False, featureless=False, **kwargs): super(GraphConvolution, self).__init__(**kwargs) if dropout: self.dropout = placeholders['dropout'] else: self.dropout = 0. self.act = act self.support = placeholders['support'] self.sparse_inputs = sparse_inputs self.featureless = featureless self.bias = bias # helper variable for sparse dropout self.num_features_nonzero = placeholders['num_features_nonzero'] # 下面是定义变量，主要是通过调用utils.py中的glorot函数实现 with tf.variable_scope(self.name + '_vars'): for i in range(len(self.support)): self.vars['weights_' + str(i)] = glorot([input_dim, output_dim], name='weights_' + str(i)) if self.bias: self.vars['bias'] = zeros([output_dim], name='bias') if self.logging: self._log_vars() def _call(self, inputs): x = inputs # dropout 设置dropout if self.sparse_inputs: x = sparse_dropout(x, 1-self.dropout, self.num_features_nonzero) else: x = tf.nn.dropout(x, 1-self.dropout) # convolve 卷积的实现。主要是根据论文中公式Z = \tilde{D}^{-1/2}\tilde{A}^{-1/2}X\theta实现 supports = list() for i in range(len(self.support)): if not self.featureless: pre_sup = dot(x, self.vars['weights_' + str(i)], sparse=self.sparse_inputs) else: pre_sup = self.vars['weights_' + str(i)] support = dot(self.support[i], pre_sup, sparse=True) supports.append(support) output = tf.add_n(supports) # bias if self.bias: output += self.vars['bias']