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

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: He, Hongyang, Song, Xinyuan, He, Yangfan, Zhang, Zeyu, Li, Yanshu, You, Haochen, Sun, Lifan, Zhang, Wenqiao
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909936319987712
author He, Hongyang
Song, Xinyuan
He, Yangfan
Zhang, Zeyu
Li, Yanshu
You, Haochen
Sun, Lifan
Zhang, Wenqiao
author_facet He, Hongyang
Song, Xinyuan
He, Yangfan
Zhang, Zeyu
Li, Yanshu
You, Haochen
Sun, Lifan
Zhang, Wenqiao
contents 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.Code:https://github.com/HoHongYeung/NeurIPS25-TRiCo.
format Preprint
id arxiv_https___arxiv_org_abs_2509_21526
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle TRiCo: Triadic Game-Theoretic Co-Training for Robust Semi-Supervised Learning
He, Hongyang
Song, Xinyuan
He, Yangfan
Zhang, Zeyu
Li, Yanshu
You, Haochen
Sun, Lifan
Zhang, Wenqiao
Machine Learning
Computer Vision and Pattern Recognition
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.Code:https://github.com/HoHongYeung/NeurIPS25-TRiCo.
title TRiCo: Triadic Game-Theoretic Co-Training for Robust Semi-Supervised Learning
topic Machine Learning
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2509.21526