Learnability with Indirect Supervision Signals

Abstract

Learning from indirect supervision signals is important in real-world AI applications when, often, gold labels are missing or too costly. In this paper, we develop a unified theoretical framework for multi-class classification when the supervision is provided by a variable that contains nonzero mutual information with the gold label. The nature of this problem is determined by (i) the transition probability from the gold labels to the indirect supervision variables and (ii) the learner's prior knowledge about the transition. Our framework relaxes assumptions made in the literature, and supports learning with unknown, non-invertible and instance-dependent transitions. Our theory introduces a novel concept called \emph{separation}, which characterizes the learnability and generalization bounds. We also demonstrate the application of our framework via concrete novel results in a variety of learning scenarios such as learning with superset annotations and joint supervision signals.

References in corpus (8)

• Training Convolutional Networks with Noisy Labels
• Deep Learning is Robust to Massive Label Noise
• Risk bounds for statistical learning
• Making Risk Minimization Tolerant to Label Noise
• Learning from Complementary Labels
• Distribution-Independent PAC Learning of Halfspaces with Massart Noise
• Estimation from Indirect Supervision with Linear Moments
• Efficient Learning of Linear Separators under Bounded Noise