RoPAWS: Robust Semi-supervised Representation Learning from Uncurated Data
This addresses the challenge of using real-world unlabeled data in semi-supervised learning for image classification, offering a robust solution for applications with uncurated datasets.
The paper tackles the problem of semi-supervised learning with uncurated data, where existing methods like PAWS underperform due to out-of-class samples; it proposes RoPAWS, which improves PAWS by +5.3% on uncurated Semi-iNat and +0.4% on curated ImageNet.
Semi-supervised learning aims to train a model using limited labels. State-of-the-art semi-supervised methods for image classification such as PAWS rely on self-supervised representations learned with large-scale unlabeled but curated data. However, PAWS is often less effective when using real-world unlabeled data that is uncurated, e.g., contains out-of-class data. We propose RoPAWS, a robust extension of PAWS that can work with real-world unlabeled data. We first reinterpret PAWS as a generative classifier that models densities using kernel density estimation. From this probabilistic perspective, we calibrate its prediction based on the densities of labeled and unlabeled data, which leads to a simple closed-form solution from the Bayes' rule. We demonstrate that RoPAWS significantly improves PAWS for uncurated Semi-iNat by +5.3% and curated ImageNet by +0.4%.