Inferring Generative Model Structure with Static Analysis
This addresses the bottleneck of labeled data for training discriminative models in machine learning, offering a novel solution for domains like radiology where ground truth is scarce.
The paper tackles the problem of learning generative model structure without ground truth labels by statically analyzing code for weak supervision sources, reducing data requirements significantly. It shows Coral improves sample complexity from exponential to quasilinear and outperforms traditional methods by up to 3.81 F1 points, even beating a fully supervised model by 3.07 accuracy points on radiology data.
Obtaining enough labeled data to robustly train complex discriminative models is a major bottleneck in the machine learning pipeline. A popular solution is combining multiple sources of weak supervision using generative models. The structure of these models affects training label quality, but is difficult to learn without any ground truth labels. We instead rely on these weak supervision sources having some structure by virtue of being encoded programmatically. We present Coral, a paradigm that infers generative model structure by statically analyzing the code for these heuristics, thus reducing the data required to learn structure significantly. We prove that Coral's sample complexity scales quasilinearly with the number of heuristics and number of relations found, improving over the standard sample complexity, which is exponential in $n$ for identifying $n^{\textrm{th}}$ degree relations. Experimentally, Coral matches or outperforms traditional structure learning approaches by up to 3.81 F1 points. Using Coral to model dependencies instead of assuming independence results in better performance than a fully supervised model by 3.07 accuracy points when heuristics are used to label radiology data without ground truth labels.