Abstract: Learning image similarity underpins many vision tasks. Discriminative metric learning seeks a class embedding. The biggest challenge is learning a metric that generalizes from training to novel, but related, test samples. It should transfer…
Abstract: Linear discriminant analysis (LDA) is a classical statistical machine-learning method that finds a linear data transformation that increases class discrimination in an optimal discriminant subspace. LDA assumes Gaussian class distributions and single-label data annotations….
Abstract: Deep learning’s success makes understanding its mechanism and rationality urgent. Deep learning requires data depth, structure, and volume. Most deep learning theoretical studies emphasize the importance and benefits of neural network depth and structure….
Abstract: Multiview fuzzy systems use multiview learning to effectively model fuzzy scenarios and obtain an interpretable model. Current multiview fuzzy system studies face several challenges, including how to efficiently collaborate between multiple views with few…
Abstract: Recently, social media research has grown. This article predicts active users in online social events, which is rare compared to previous work on emerging topics. This prediction task is a binary classification problem based…
Abstract: In various domains and settings, model-based approaches have produced effective recommendation lists. They can accurately recommend items using latent factors. Even if the model embeds content-based information, we lose references to the recommended item’s…
Abstract: Kernel methods have thrived for 20 years. Big data has greatly increased data collection. Existing kernel methods are not scalable for training or predicting. This paper introduces a general sparse kernel learning formulation based…
Abstract: The curse of kernelization affects most kernel-based models, including SVM, despite their strong generalization ability. Hyperparameter tuning, which requires computational resources, affects their predictive performance. Kernel methods have these issues with large datasets. We…
Abstract: Background clutter often affects RGBT tracking methods that localize a target object with a bounding box. This article introduces noise-robust cross-modal ranking to reduce background effects in RGBT target bounding boxes. We address noise…
Abstract: Machine learning models have excelled in machine translation, face recognition, and recommendation thanks to representation learning. Most representation learning methods require many consistent and noise-free labels. Labels are often limited due to budget constraints…
Abstract: Modern semantic segmentation methods require labeled data and require fine-tuning to work on unseen classes. Thus, few-shot segmentation proposes learning a model that quickly adapts to new classes with a few labeled support samples….
Abstract: Geographical topic models are used in recommendations, user mobility modeling, event detection, and geo-tagged document mining. Studying effective geographical topic models ignores efficiency. It takes days to train a geographical topic model on a…
WhatsApp us