Node Clustering Scheme

• is an extension of the exact inference algorithm
• idea: joins nodes together before feeding them to an exact inference algorithm
• lays the foundation to more generalized methods:
  • Junction Graph Belief Propagation
  • Junction Tree Belief Propagation

Node Clustering - Examples

Click here to expand...

Example 1

Example 2

Node Clustering - Limitations

• if two variables are to be joined, 𝐶 having |𝐶| states and 𝐷 having |𝐷| states the new variable 𝐶𝐷 will have |𝐶|*|𝐷| states
• the number of states in joined variables goes up exponentially

Node Clustering - Generalized Approaches

• Junction Graph Belief Propagation
• Junction Tree Belief Propagation