Active learning aims to reduce annotation cost by predicting which samples are useful for a human teacher to label. However it has become clear there is no best active learning algorithm. Inspired by various philosophies about what constitutes a good criteria, different algorithms perform well on different datasets. This has motivated research into ensembles of active learners that learn what constitutes a good criteria in a given scenario, typically via multi-armed bandit algorithms. Though algorithm ensembles can lead to better results, they overlook the fact that not only does algorithm efficacy vary across datasets, but also during a single active learning session. That is, the best criteria is non-stationary. This breaks existing algorithms' guarantees and hampers their performance in practice. In this paper, we propose dynamic ensemble active learning as a more general and promising research direction. We develop a dynamic ensemble active learner based on a non-stationary multi-armed bandit with expert advice algorithm. Our dynamic ensemble selects the right criteria at each step of active learning. It has theoretical guarantees, and shows encouraging results on 13 popular datasets.

No abstract

Hyperspectral imaging combined with chemometric approaches is proven to be a powerful tool for the quality evaluation and control of fruits. In fruit defect-detection scenarios, developing an unsupervised anomaly detection framework is vital, as defect sample preparation is labor-intensive and time-consuming, especially for exploring potential defects. In this paper, a spectral–spatial, information-based, self-supervised anomaly detection (SSAD) approach is proposed. During training, an auxiliary classifier is proposed to identify the projection axes of principal component (PC) images that were transformed from the hyperspectral data cubes. In test time, the fully connected layer of the learned classifier was used as a ‘spectral–spatial’ feature extractor, and the feature similarity metric was adopted as the score function for the downstream anomaly evaluation task. The proposed network was evaluated with two fruit data sets: a strawberry data set with bruised, infected, chilling-injured, and contaminated test samples and a blueberry data set with bruised, infected, chilling-injured, and wrinkled samples as anomalies. The results show that the SSAD yielded the best anomaly detection performance (AUC = 0.923 on average) over the baseline methods, and the visualization results further confirmed its advantage in extracting effective ‘spectral–spatial’ latent representation. Moreover, the robustness of SSAD is verified with the data pollution experiment; it performed significantly better than the baselines when a portion of anomalous samples was involved in the training process.

Automatic gesture synthesis from speech is a topic that has attracted researchers for applications in remote communication, video games and Metaverse. Learning the mapping between speech and 3D full-body gestures is difficult due to the stochastic nature of the problem and the lack of a rich cross-modal dataset that is needed for training. In this paper, we propose a novel transformer-based framework for automatic 3D body gesture synthesis from speech. To learn the stochastic nature of the body gesture during speech, we propose a variational transformer to effectively model a probabilistic distribution over gestures, which can produce diverse gestures during inference. Furthermore, we introduce a mode positional embedding layer to capture the different motion speeds in different speaking modes. To cope with the scarcity of data, we design an intra-modal pre-training scheme that can learn the complex mapping between the speech and the 3D gesture from a limited amount of data. Our system is trained with either the Trinity speech-gesture dataset or the Talking With Hands 16.2M dataset. The results show that our system can produce more realistic, appropriate, and diverse body gestures compared to existing state-of-the-art approaches.

No abstract