Discriminative Learning of Bayesian Networks via Factorized Conditional Log-Likelihood
See where this sits in the topic map →Summary AI-generated
- TL;DR
- We propose an efficient and parameter-free scoring criterion called factorized conditional log-likelihood (fCLL) for learning Bayesian network classifiers.
- Problem
- Not specified in the abstract.
- Method
- The method approximates the conditional log-likelihood criterion to guarantee decomposability over the network structure and efficient parameter estimation, matching the time and space complexity of traditional log-likelihood scoring. It also features an information-theoretic interpretation based on interaction information.
- Results
- Experiments across a large suite of UCI benchmark datasets show that fCLL-trained classifiers achieve accuracy matching the best compared classifiers while using significantly fewer computational resources.
- Contributions
- Not specified in the abstract.
- Limitations
- Not specified in the abstract.
- Takeaways
- The fCLL scoring criterion enables efficient, parameter-free discriminative learning for Bayesian network classifiers without sacrificing classification accuracy.
- Applications
- Not specified in the abstract.
- Topics
- Discriminative Learning; Bayesian Networks; Machine Learning
- For industry
- Not specified in the abstract.
- Why it matters
- Not specified in the abstract.
Abstract
We propose an efficient and parameter-free scoring criterion, the factorized conditional log-likelihood (ˆfCLL), for learning Bayesian network classifiers. The proposed score is an approximation of the conditional log-likelihood criterion. The approximation is devised in order to guarantee decomposability over the network structure, as well as efficient estimation of the optimal parameters, achieving the same time and space complexity as the traditional log-likelihood scoring criterion. The resulting criterion has an information-theoretic interpretation based on interaction information, which exhibits its discriminative nature. To evaluate the performance of the proposed criterion, we present an empirical comparison with state-of-the-art classifiers. Results on a large suite of benchmark data sets from the UCI repository show that ˆfCLL-trained classifiers achieve at least as good accuracy as the best compared classifiers, using significantly less computational resources.