Active learning (AL) seeks to reduce annotation costs by selecting the most informative samples for labeling, making it particularly valuable in resource-constrained settings. However, traditional evaluation methods, which focus solely on final accuracy, fail to capture the full dynamics of the learning process. To address this gap, we propose PALM (Performance Analysis of Active Learning Models), a unified and interpretable mathematical model that characterizes AL trajectories through four key parameters: achievable accuracy, coverage efficiency, early-stage performance, and scalability. PALM provides a predictive description of AL behavior from partial observations, enabling the estimation of future performance and facilitating principled comparisons across different strategies. We validate PALM through extensive experiments on CIFAR-10/100 and ImageNet-50/100/200, covering a wide range of AL methods and self-supervised embeddings. Our results demonstrate that PALM generalizes effectively across datasets, budgets, and strategies, accurately predicting full learning curves from limited labeled data. Importantly, PALM reveals crucial insights into learning efficiency, data space coverage, and the scalability of AL methods. By enabling the selection of cost-effective strategies and predicting performance under tight budget constraints, PALM lays the basis for more systematic, reproducible, and data-efficient evaluation of AL in both research and real-world applications. The code is available at: https://github.com/juliamachnio/PALM.

Learning temporal patterns from multivariate longitudinal data is challenging especially in cases when data is sporadic, as often seen in, e.g., healthcare applications where the data can suffer from irregularity and asynchronicity as the time between consecutive data points can vary across features and samples, hindering the application of existing deep learning models that are constructed for complete, evenly spaced data with fixed sequence lengths. In this article, a novel deep learning-based model is developed for modeling multiple temporal features in sporadic data using an integrated deep learning architecture based on a recurrent neural network (RNN) unit and a continuous-time autoregressive (CAR) model. The proposed model, called CARRNN, uses a generalized discrete-time autoregressive (AR) model that is trainable end-to-end using neural networks modulated by time lags to describe the changes caused by the irregularity and asynchronicity. It is applied to time-series regression and classification tasks for Alzheimer's disease progression modeling, intensive care unit (ICU) mortality rate prediction, human activity recognition, and event-based digit recognition, where the proposed model based on a gated recurrent unit (GRU) in all cases achieves significantly better predictive performance than the state-of-the-art methods using RNNs, GRUs, and long short-term memory (LSTM) networks.

Automatic emotion recognition in the wild video datasets is a very challenging problem because of the inter-class similarities among different facial expressions and large intra-class variabilities due to the significant changes in illumination, pose, scene, and expression. In this paper, we present our proposed method for video-based emotion recognition in the EmotiW 2016 challenge. The task considers the unconstrained emotion recognition problem by training from short video clips extracted from movies and testing on short movie clips and spontaneous video clips of the reality TV data. Four different methods are employed to extract both static and dynamic emotion representations from the videos. First, local binary patterns of three orthogonal planes are used to describe spatiotemporal features of the video frames. Second, principal component analysis is applied to the image patches in a two-stage convolutional network to learn weights and extract facial features from the aligned faces. Third, the deep convolutional neural network model of VGG-Face is deployed to extract deep facial representations from aligned faces. Fourth, a bag of visual words is computed based on dense scale-invariant feature transform descriptors from aligned face images to form hand-crafted representations. Support vector machines are then utilized to train and classify the obtained spatiotemporal representations and facial features. Finally, score-level fusion is applied to combine the classification results and predict the emotion labels of the video clips. The results show that the proposed combined method has outperformed all the utilized techniques with the overall validations and test accuracies of 43.13% and 40.13%, respectively. This system, is relatively a good classifier in Happy and Angry emotion categories and is unsuccessful in detecting Surprise, Disgust, and Fear.