Abstract:Active learning is an effective method for iteratively selecting a subset of images from an unlabeled dataset. One of the most widely used active learning strategies is uncertainty sampling. However, traditional sampling strategies do not take the category of samples into consideration, and the selected images do not reflect the desired training distribution, leading to the result that additional labeling work needs to be done. To deal with these problems, from the aspect of visual perception, we improve the traditional entropy-based uncertainty sampling strategy by introducing a certainty measurement estimated by a bag-of-visual-words (BoVW). The Rescorla–Wagner perceptive model is utilized to quantify the stop criterion. This method differs from previous approaches that treated sampling and classifying process separately: we treat the learning process as a uniform model by proposing a new evolving sample selection method that uses the unified negative-accelerated learning principle and takes category distribution into consideration. A classifier is trained to provide category distributions for the sampling process to improve its sampling performance and reduce additional annotation costs for the human annotator. During the training process, weights for both modules are adaptively initialized by the inner similarity of sample set measured by structural similarity (SSIM), and dynamically adjusted according to the learning process of the human. In addition to the regular tests that are commonly utilized by traditional sampling methods, the transfer test, based on transfer learning theory, is utilized to further evaluate the performance of different sampling strategies. Experimental results on real world datasets show that our active sampling framework outperforms both baseline and state-of-the-art adaptive active learning strategies.

Active learning with adaptive regularization

Uncertainty-aware Complementary Label Queries for Active Learning

Uncertainty Sampling Based Active Learning with Diversity Constraint by Sparse Selection.

Uncertainty-Based Active Learning Via Sparse Modeling for Image Classification

Hierarchical exploration based active learning with support vector machine

A Serial Sample Selection Framework for Active Learning.

Learning Distinctive Margin Toward Active Domain Adaptation

Efficient Active Learning by Querying Discriminative and Representative Samples and Fully Exploiting Unlabeled Data

Active Learning Via Sequential Design and Uncertainty Sampling

A Multicriterion Query-Based Batch Mode Active Learning Technique

Adaptive Informative Sampling for Active Learning.

Cost-sensitive Active Learning with a Label Uniform Distribution Model

ASCENT: Active Supervision for Semi-Supervised Learning.

Active learning for binary classification with variable selection

Improved Adaptive Algorithm for Scalable Active Learning with Weak Labeler

Uncertainty-aware Active Learning for Optimal Bayesian Classifier

K-means Based on Active Learning for Support Vector Machine.

Practice Makes Perfect: an Adaptive Active Learning Framework for Image Classification

Active Learning Based on Belief Functions

Active Learning with Adaptive Heterogeneous Ensembles

Asymmetric semi-supervised boosting for SVM active learning in CBIR