Quantifying Sensitivity for Tree Ensembles: A symbolic and compositional approach
Researchers have developed a new algorithmic technique to quantify the sensitivity of decision tree ensembles (DTEs). This method discretizes the input space to identify regions susceptible to misclassification from small feature changes. The approach, which encodes the problem using algebraic decision diagrams, offers a compositional and scalable solution that has demonstrated significant speedups over existing methods in experiments. AI
IMPACT Provides a more efficient and scalable method for verifying properties of decision tree ensembles, crucial for safety-critical AI applications.