Online semi-supervised perception: Real-time learning without explicit feedback

2026-04-30

AI Summary

This work addresses the challenge of dynamic perception tasks where unlabeled data arrive continuously yet explicit feedback is absent. It proposes the first online semi-supervised learning framework capable of real-time model updates. The approach integrates graph-structured modeling with an incremental graph-update mechanism, initializing with a small set of offline labeled samples to establish a prior bias and subsequently refining the graph structure dynamically using streaming unlabeled data to enable continual learning. Theoretical analysis provides regret bounds that guarantee the algorithm’s performance. Experimental results on three video datasets demonstrate that the framework achieves high precision and recall in face recognition tasks while supporting real-time inference.

Abstract

This paper proposes an algorithm for real-time learning without explicit feedback. The algorithm combines the ideas of semi-supervised learning on graphs and online learning. In particular, it iteratively builds a graphical representation of its world and updates it with observed examples. Labeled examples constitute the initial bias of the algorithm and are provided offline, and a stream of unlabeled examples is collected online to update this bias. We motivate the algorithm, discuss how to implement it efficiently, prove a regret bound on the quality of its solutions, and apply it to the problem of real-time face recognition. Our recognizer runs in real time, and achieves superior precision and recall on 3 challenging video datasets.

Problem

• online semi-supervised learning
• real-time learning
• no explicit feedback
• face recognition
• graph-based learning

Innovation

• online semi-supervised learning
• graph-based learning
• real-time perception
• regret bound
• face recognition