RealMix: Towards Realistic Semi-Supervised Deep Learning Algorithms

18 Dec 2019  ·  Varun Nair, Javier Fuentes Alonso, Tony Beltramelli ·

Semi-Supervised Learning (SSL) algorithms have shown great potential in training regimes when access to labeled data is scarce but access to unlabeled data is plentiful. However, our experiments illustrate several shortcomings that prior SSL algorithms suffer from. In particular, poor performance when unlabeled and labeled data distributions differ. To address these observations, we develop RealMix, which achieves state-of-the-art results on standard benchmark datasets across different labeled and unlabeled set sizes while overcoming the aforementioned challenges. Notably, RealMix achieves an error rate of 9.79% on CIFAR10 with 250 labels and is the only SSL method tested able to surpass baseline performance when there is significant mismatch in the labeled and unlabeled data distributions. RealMix demonstrates how SSL can be used in real world situations with limited access to both data and compute and guides further research in SSL with practical applicability in mind.

PDF Abstract

Datasets


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Semi-Supervised Image Classification cifar10, 250 Labels RealMix Percentage correct 90.21 # 3
Semi-Supervised Image Classification CIFAR-10, 250 Labels RealMix Percentage error 9.79 # 17
Semi-Supervised Image Classification CIFAR-10, 250 Labels EnAET Percentage error 7.6 # 16
Semi-Supervised Image Classification CIFAR-10, 4000 Labels RealMix Percentage error 6.38 # 28
Semi-Supervised Image Classification SVHN, 250 Labels RealMix Accuracy 96.47 # 7

Methods


No methods listed for this paper. Add relevant methods here