Regularized Data Programming with Automated Bayesian Prior Selection

TL;DR

This paper introduces a Bayesian regularization approach to data programming that automates prior selection, improving dataset labeling accuracy especially in low-data scenarios, and enhances interpretability over traditional methods.

Contribution

It proposes a Bayesian extension of data programming with automated prior selection using majority vote as a proxy, addressing failures of unsupervised learning.

Findings

Regularized DP outperforms maximum likelihood and majority voting.

Improves performance in low-data regimes.

Enhances interpretability of the labeling process.

Abstract

The cost of manual data labeling can be a significant obstacle in supervised learning. Data programming (DP) offers a weakly supervised solution for training dataset creation, wherein the outputs of user-defined programmatic labeling functions (LFs) are reconciled through unsupervised learning. However, DP can fail to outperform an unweighted majority vote in some scenarios, including low-data contexts. This work introduces a Bayesian extension of classical DP that mitigates failures of unsupervised learning by augmenting the DP objective with regularization terms. Regularized learning is achieved through maximum a posteriori estimation with informative priors. Majority vote is proposed as a proxy signal for automated prior parameter selection. Results suggest that regularized DP improves performance relative to maximum likelihood and majority voting, confers greater interpretability, and bolsters performance in low-data regimes.