As I have indicated in previous messages in the Help forum, we are currently designing high-level interfaces that will wrap the existing PNL API into more-or-less flat versions that will allow access to most of the basic functionality that will be used for Bayesian networks. We are currently designing and implementing a wrapper for BNs only, but we will also continue with DBNs, MRFs and Factor graphs later.
These wrappers will handle memory allocation and will also manage a lot of the data-structure-consistency that is required by the existing API (for example nodes/arcs could be added incrementally rather than in batch and no topological ordering will be required).
The question is, what functionality should be available at this high level?
Our current list looks something like this:
Read/Write network files
Read/Write Data Files (for learning)
Incremental Add/Del Nodes
Incremental Add/Del Arcs
Query the network for relevant information such as structural information and CPD info
Sample Data from the Network
Attach User-specific data to nodes/network
Are there any other features that you are using routinely that is not included in this list?
Also, what additional features would you like to see for DBNs, MRFs and Factor Graphs?
Log in to post a comment.