Targeting Negative Flips in Active Learning using Validation Sets

TL;DR

ROSE - a plug-in algorithm that utilizes a small labeled validation set to restrict arbitrary active learning acquisition functions to negative flips within the unlabeled pool and shows that integrating a validation set results in a significant performance boost in terms of accuracy, negative flip rate reduction, or both.

Abstract

The performance of active learning algorithms can be improved in two ways. The often used and intuitive way is by reducing the overall error rate within the test set. The second way is to ensure that correct predictions are not forgotten when the training set is increased in between rounds. The former is measured by the accuracy of the model and the latter is captured in negative flips between rounds. Negative flips are samples that are correctly predicted when trained with the previous/smaller dataset and incorrectly predicted after additional samples are labeled. In this paper, we discuss improving the performance of active learning algorithms both in terms of prediction accuracy and negative flips. The first observation we make in this paper is that negative flips and overall error rates are decoupled and reducing one does not necessarily imply that the other is reduced. Our observation is important as current active learning algorithms do not consider negative flips directly and implicitly assume the opposite. The second observation is that performing targeted active learning on subsets of the unlabeled pool has a significant impact on the behavior of the active learning algorithm and influences both negative flips and prediction accuracy. We then develop ROSE - a plug-in algorithm that utilizes a small labeled validation set to restrict arbitrary active learning acquisition functions to negative flips within the unlabeled pool. We show that integrating a validation set results in a significant performance boost in terms of accuracy, negative flip rate reduction, or both.

Figures (9)

• Figure 1: Overview of our method, RoSE, and learning curves of Entropy sampling on CINIC10. a) high-level diagramm of RoSE. b) accuracy and negative flip rate for the CINIC10 dataset. Higher accuracy and lower negative flip rate values are better.
• Figure 2: Accuracy and NFR curves of three different acquisition functions on the CIFAR10 and CIFAR100 benchmark. The plots consider entropy sampling, margin sampling, and least confidence sampling. For accuracy, higher values indicates strong performance. For NFR, lower values are preferred. a) CIFAR100 accuracy and NFR performance. b) CIFAR10 accuracy and NFR performance.
• Figure 3: Class complexity analysis of the NFR. Each plot represents a different number of classes from CIFAR100. a) 5 classes; b) 50 classes.
• Figure 4: NFR on 15 classes in the CIFAR100 dataset. a) balanced class setting for both unlabeled pool and test set. b) imbalanced setting.
• Figure 5: Random sampling from the restricted unlabeled pool based on the subsets both correct (BC), both wrong (BW), negative flips (NF), and positive flips (PF). a) Accuracy; b) NFR.