ProML: Progressive Multi-label Learning

Key points

• Multi-label learning typically pursues model construction through the joint optimization of label fitting, correlation modeling, and structural regularization.1
• While empirically effective, such a coupled paradigm imposes multiple learning objectives on a single predictive parameter matrix, forcing it to simultaneously fit observed labels and encode structural priors.1
• We theoretically analyze the optimization geometry of ProML and show that, under quadratic positive semidefinite structural regularization, ProML attenuates the dominant curvature of the fitting term, enlarges the stable learning rate range, and provides a controllable fidelity-regularization trade-off.1
• Further, we instantiate the ProML for multi-label feature selection task by integrating global-local label priors and row-sparse regularization.1

Sources (1)

• [1]ProML: Progressive Multi-label Learning
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 11, 01:37 PM

  To address the these issues, this paper proposes a Progressive Multi-Label Learning (ProML) framework, which decouples direct label fitting from predictor learning via the introduction of an auxiliary predictive variable.

  Multi-label learning typically pursues model construction through the joint optimization of label fitting, correlation modeling, and structural regularization.

Extractive summary: sentences quoted from the sources.