LiDAM: Semi-Supervised Learning with Localized Domain Adaptation and Iterative Matching

13 Oct 2020  ·  Qun Liu, Matthew Shreve, Raja Bala ·

Although data is abundant, data labeling is expensive. Semi-supervised learning methods combine a few labeled samples with a large corpus of unlabeled data to effectively train models. This paper introduces our proposed method LiDAM, a semi-supervised learning approach rooted in both domain adaptation and self-paced learning. LiDAM first performs localized domain shifts to extract better domain-invariant features for the model that results in more accurate clusters and pseudo-labels. These pseudo-labels are then aligned with real class labels in a self-paced fashion using a novel iterative matching technique that is based on majority consistency over high-confidence predictions. Simultaneously, a final classifier is trained to predict ground-truth labels until convergence. LiDAM achieves state-of-the-art performance on the CIFAR-100 dataset, outperforming FixMatch (73.50% vs. 71.82%) when using 2500 labels.

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Benchmark
Semi-Supervised Image Classification cifar-100, 10000 Labels LiDAM Percentage error 23.22 # 18
Semi-Supervised Image Classification CIFAR-100, 2500 Labels LiDAM Percentage error 26.50 # 8
Semi-Supervised Image Classification CIFAR-100, 5000Labels LiDAM Percentage correct 75.14 # 1
Semi-Supervised Image Classification CIFAR-100, 5000 Labels LiDAM Accuracy (%) 75.14 # 1
Semi-Supervised Image Classification CIFAR-10, 1000 Labels LiDAM Accuracy 89.04 # 4
Semi-Supervised Image Classification CIFAR-10, 250 Labels LiDAM Percentage error 19.17 # 19
Semi-Supervised Image Classification CIFAR-10, 4000 Labels LiDAM Percentage error 7.48 # 32

Methods