DST

AI
2026-07-22 07:54:07

ICML 2026 paper says random noise can serve as a transfer source in low-label learning

A study presented at ICML 2026 argues that transfer learning does not always need a semantically meaningful source domain. The paper, titled "Semi-Supervised Noise Adaptation: Transferring Knowledge from Noise Domain," introduces Semi-Supervised Noise Adaptation (SSNA) and a Noise Adaptation Framework (NAF) that builds class-separable structure from randomly sampled Gaussian noise, then aligns that structure with a target domain using a small number of labeled examples. Under a 4-labels-per-class setting and a ResNet-18 backbone, NAF outperformed the standard empirical risk minimization baseline on CIFAR-10, CIFAR-100, DTD-47, and Caltech-101 by 12.35, 7.61, 4.38, and 2.74 percentage points, respectively. The study also reported gains on fine-grained datasets, ImageNet-1K, and the AG News-4 text classification task with BERT. The code has been open-sourced. Ablation results in the paper suggest the useful part is not randomness itself, but whether the noise domain forms a separable class structure in representation space. When that structure collapses, performance falls sharply. When class centers are pulled farther apart, results improve. The work frames synthetic noise as a low-cost substitute when real source data cannot be shared because of privacy, confidentiality, or copyright limits.

460
ICML 2026 paper says random noise can serve as a transfer source in low-label learning