Exact Inference

Context

• By default, BayesiaLab constructs a Junction Tree from the given Bayesian network to perform Exact Inference.
• BayesiaLab constructs a Junction Tree as you switch from Modeling Mode F4 to Validation Mode F5 for a given Bayesian network model for the first time.
• Constructing such a Junction Tree can be very time and memory-consuming depending on the network size and complexity.
• However, once the Junction Tree is constructed, performing inference is very quick.
• Whenever you modify your network, i.e., structure or parameters, BayesiaLab needs to recreate the entire Junction Tree again.

Usage

• You can specify to use Exact Inference by selecting Menu > Inference > Exact Inference.

• By default, Exact Inference is active whenever you build or learn a new Bayesian network.
• In certain cases, the memory requirements or the expected computing time for creating the Junction Tree can be prohibitive.
• If you switch from Modeling Mode F4 to Validation Mode F5 in such a situation, BayesiaLab may display a warning and offer you several options depending on the severity of the situation:

Condition	Exact Inference will be very time-consuming	Exact Inference will not be possible
Options
Continue with Exact Inference	- You can still proceed with Exact Inference despite the warning. - However, you may face a long wait or encounter system stability issues.	n/a
Use the Automatic Structural Complexity Reducer	- You can take advantage of the Automatic Complexity Reducer, which removes less important arcs in the network. - To do so, the Automatic Complexity Reducer uses the dataset associated with the current network — or generates one according to the probability distributions — to compute the importance of each arc in the network. - The Automatic Complexity Reducer deletes the least important arcs until Exact Inference becomes possible in terms of memory and computation time. - The Automatic Complexity Reducer concludes its process by displaying a report listing all arcs that were deleted.	- You can take advantage of the Automatic Complexity Reducer, which removes less important arcs in the network. - To do so, the Automatic Complexity Reducer uses the dataset associated with the current network — or generates one according to the probability distributions — to compute the importance of each arc in the network. - The Automatic Complexity Reducer deletes the least important arcs until Exact Inference becomes possible in terms of memory and computation time. - The Automatic Complexity Reducer concludes its process by displaying a report listing all arcs that were deleted.
Go back to Modeling Mode	- You can return to the Modeling Mode F4 so you can modify the network yourself.	n/a
Switch to Approximate Inference	- Using Approximate Inference avoids the memory problem but sacrifices inference precision. Also, many types of analyses in BayesiaLab only work with Exact Inference.	- Using Approximate Inference avoids the memory problem but sacrifices inference precision. Also, many types of analyses in BayesiaLab only work with Exact Inference.