UCLA Automated Reasoning Group
Part of a talk given at PKU, which discusses an approach for explaining and verifying Bayesian Network classifiers. First, the classifier is compiled into a symbolic decision graph. Next, classical AI and CS techniques are applied to the decision graph, to explain the classifier’s decisions and to formally verify some of its properties.
This work combines techniques from Eras 1 & 2 in the history of AI; see the talk “On AI Education” at https://youtu.be/bzaMyYqJ030
Links:
A Symbolic Approach to Explaining Bayesian Network Classifiers (IJCAI-18)
http://reasoning.cs.ucla.edu/fetch.php?id=180&type=pdf
Formal Verification of Bayesian Network Classifiers (PGM-18)
http://reasoning.cs.ucla.edu/fetch.php?id=182&type=pdf
Reasoning About Bayesian Network Classifiers (UAI-03)
http://reasoning.cs.ucla.edu/fetch.php?id=15&type=pdf
Source
