add l2c description

docs/algorithms.md

@@ -11,6 +11,6 @@ Here we are comparing the different algorithms deployed.
 | 6    | Torus     | A torus network is a grid network with "wrap-around" connections at the edges. This allows representations to exit from one edge and re-enter from the opposite edge, creating a continuous path. Routing algorithms for torus, leverage similar principles as grid routing while accounting for the wrap-around connections. |            |
 | 7    | Similarity based (top-k) | Similarity-based (top-k) algorithms utilize similarity measures to compare the data packet's destination with the characteristics of neighboring nodes. Nodes with higher similarity to the destination are more likely to be chosen as the next hop. This can be effective in content-based routing scenarios where representations aim to reach nodes with specific content or properties. |            |
 | 8    | Swarm     | Nodes in the network communicate and share information about their local routing experiences. This collaborative approach can lead to adaptive and efficient routing, particularly in dynamic networks. This is inspired by the collective behavior of biological swarms. |            |
-| 9    | L2C       | Description of the L2C algorithm |            |
+| 9    | L2C       | L2C algorithms use a model encoder to learn collaboration weights that optimize performance on a local validation set. Nodes dynamically adjust their collaboration strategies based on these learned weights, allowing for more effective peer-to-peer communication. This approach is particularly useful in decentralized networks where data heterogeneity is significant, as it enables nodes to prioritize collaboration with those whose data distributions are more aligned with their own, leading to improved convergence and overall performance. |            |