TRiCo: Triadic Game-Theoretic Co-Training for Robust Semi-Supervised Learning

2025-09-25

AI Summary

Existing semi-supervised learning (SSL) methods suffer from static view interaction, unreliable pseudo-labels, and insufficient modeling of hard samples. To address these limitations, we propose TRiCo, a tripartite Stackelberg game framework involving a meta-teacher, two student classifiers, and an adversarial generator—enabling dynamic cross-view collaboration and robust pseudo-label generation. Our approach introduces a mutual-information-driven uncertainty estimation mechanism for adaptive pseudo-label selection and loss weighting. We explicitly model decision-boundary vulnerabilities via frozen-representation dual students, a meta-learned teacher, and a non-parametric embedding perturbation generator. TRiCo achieves state-of-the-art performance on CIFAR-10, SVHN, STL-10, and low-label ImageNet benchmarks. It is architecture-agnostic and compatible with frozen vision backbones, significantly improving generalization and robustness under limited supervision.

Technology Category

• Machine Learning: Semi-Supervised Learning
• Computer Vision: Adversarial Attacks & Robustness
• Game Theory and Economic Paradigms: Adversarial Learning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI

Abstract

We introduce TRiCo, a novel triadic game-theoretic co-training framework that rethinks the structure of semi-supervised learning by incorporating a teacher, two students, and an adversarial generator into a unified training paradigm. Unlike existing co-training or teacher-student approaches, TRiCo formulates SSL as a structured interaction among three roles: (i) two student classifiers trained on frozen, complementary representations, (ii) a meta-learned teacher that adaptively regulates pseudo-label selection and loss balancing via validation-based feedback, and (iii) a non-parametric generator that perturbs embeddings to uncover decision boundary weaknesses. Pseudo-labels are selected based on mutual information rather than confidence, providing a more robust measure of epistemic uncertainty. This triadic interaction is formalized as a Stackelberg game, where the teacher leads strategy optimization and students follow under adversarial perturbations. By addressing key limitations in existing SSL frameworks, such as static view interactions, unreliable pseudo-labels, and lack of hard sample modeling, TRiCo provides a principled and generalizable solution. Extensive experiments on CIFAR-10, SVHN, STL-10, and ImageNet demonstrate that TRiCo consistently achieves state-of-the-art performance in low-label regimes, while remaining architecture-agnostic and compatible with frozen vision backbones.

Problem

• Rethinking semi-supervised learning with triadic game-theoretic interactions
• Addressing unreliable pseudo-labels and static view limitations
• Improving robustness in low-label regimes through adversarial boundary exploration

Innovation

• Triadic game-theoretic co-training with three interactive roles
• Meta-learned teacher adaptively regulates pseudo-label selection
• Non-parametric generator perturbs embeddings to expose weaknesses