Bayes Net Collaborative AI Research Web ToolWhatWhyHowBayes NetCollaborativeAI ResearchWeb ToolWhat•Bayesian Network = DAG that models a system•Node = variable, edge = interrelation•Each node has a local distribution•Paths are taken through the network following the flow of edges•These paths build up fancy equations•Solving these equations reveals probabilistic relationships•We need:•A.) A graphical tool to convey and manipulate networks•B.) A way to track network evaluations over time and•C.) A way for multiple users to collaborate over multiple networksWhy•Graphical representations of bayes nets can be used to teach AI•Bayesian networks can be very visual, thus lending to a hands-on experience for students•Online software can allow researchers to collaborate•Different algorithms, networks, and data histories can be shared amongst manyHow•Starting from a bayes net GUI tool with core functionality, we can add:–Further modeling features–Extra evaluation tools–A suite of evaluation algorithms•Simulated annealing•Sum-product•Pearl’s algorithm•Etc.•The enhanced GUI tool can be extended into a web service, sitting on top of a database containing:–User profiles–Network evaluation histories–Personalized network
View Full Document