anew

_posts/2024-05-01-cognn.md

@@ -10,20 +10,20 @@ related_posts: false
 #toc: true
 ---
 
-## Detailed Analysis of Cooperative Graph Neural Networks (Co-GNNs)
+## Cooperative Graph Neural Networks (Co-GNNs)
 
 - Proposed by [Finkelstein et al.](https://arxiv.org/abs/2310.01267)
 
 ### **Framework Overview**
 
 Co-GNNs introduce a novel, flexible message-passing mechanism where each node in the graph dynamically selects from the actions: `listen`, `broadcast`, `listen and broadcast`, or `isolate`. This is facilitated by two cooperating networks:
 
-1. **Action Network \(\pi\)**: Determines the optimal action for each node.
-2. **Environment Network \(\eta\)**: Updates the node states based on the chosen actions.
+1. **Action Network \( \pi \)**: Determines the optimal action for each node.
+2. **Environment Network \( \eta \)**: Updates the node states based on the chosen actions.
 
 ## **Mathematical Formulation**
 
-1. **Action Selection (Action Network \(\pi\))**:
+1. **Action Selection (Action Network \( \pi \))**:
 
    - For each node \(v\), the action network predicts a probability distribution \(p^{(\ell)}_v\) over the actions \(\{S, L, B, I\}\) at layer \(\ell\):