Improving OOD Generalization of Pre-trained Encoders via Aligned Embedding-Space Ensembles
This addresses a bottleneck in unsupervised learning for improving OOD generalization, but it is incremental as it builds on existing ensemble methods.
The paper tackled the problem of poor out-of-distribution generalization in self-supervised pre-trained embeddings by developing an unsupervised method to align embedding spaces in ensembles, resulting in improved embedding quality on both in-distribution and OOD data for MNIST.
The quality of self-supervised pre-trained embeddings on out-of-distribution (OOD) data is poor without fine-tuning. A straightforward and simple approach to improving the generalization of pre-trained representation to OOD data is the use of deep ensembles. However, obtaining an effective ensemble in the embedding space with only unlabeled data remains an unsolved problem. We first perform a theoretical analysis that reveals the relationship between individual hyperspherical embedding spaces in an ensemble. We then design a principled method to align these embedding spaces in an unsupervised manner. Experimental results on the MNIST dataset show that our embedding-space ensemble method improves pre-trained embedding quality on in-distribution and OOD data compared to single encoders.