Fix GATConv documentation (#9360)

torch_geometric/nn/conv/gat_conv.py

@@ -37,9 +37,8 @@ class GATConv(MessagePassing):
     <https://arxiv.org/abs/1710.10903>`_ paper.
 
     .. math::
-        \mathbf{x}^{\prime}_i = \alpha_{i,i}\mathbf{\Theta}_{s}\mathbf{x}_{i} +
-        \sum_{j \in \mathcal{N}(i)}
-        \alpha_{i,j}\mathbf{\Theta}_{t}\mathbf{x}_{j},
+        \mathbf{x}^{\prime}_i = \sum_{j \in \mathcal{N}(i) \cup \{ i \}}
+        \alpha_{i,j}\mathbf{\Theta}_t\mathbf{x}_{j},
 
     where the attention coefficients :math:`\alpha_{i,j}` are computed as
 

torch_geometric/nn/conv/gatv2_conv.py

@@ -41,8 +41,7 @@ class GATv2Conv(MessagePassing):
     In contrast, in :class:`GATv2`, every node can attend to any other node.
 
     .. math::
-        \mathbf{x}^{\prime}_i = \alpha_{i,i}\mathbf{\Theta}_{s}\mathbf{x}_{i} +
-        \sum_{j \in \mathcal{N}(i)}
+        \mathbf{x}^{\prime}_i = \sum_{j \in \mathcal{N}(i) \cup \{ i \}}
         \alpha_{i,j}\mathbf{\Theta}_{t}\mathbf{x}_{j},
 
     where the attention coefficients :math:`\alpha_{i,j}` are computed as

torch_geometric/nn/models/lightgcn.py

@@ -275,7 +275,7 @@ class BPRLoss(_Loss):
         \sum_{j \not\in \mathcal{N}_u} \ln \sigma(\hat{y}_{ui} - \hat{y}_{uj})
         + \lambda \vert\vert \textbf{x}^{(0)} \vert\vert^2
 
-    where :math:`lambda` controls the :math:`L_2` regularization strength.
+    where :math:`\lambda` controls the :math:`L_2` regularization strength.
     We compute the mean BPR loss for simplicity.
 
     Args: