Abstract: GPS and wireless technology generate massive trajectory data. LBSN activity trajectory adds user semantic activities like visiting work/home/entertainment places to traditional trajectory data. Comparing activity trajectories in time, location, and semantics measures their similarity….
Abstract: Sparse representation-based image classification is a key component of modern computer vision systems due to its robustness. On sparse learning system optimization, regularization and data augmentation are powerful but isolated. Regularization and data augmentation…
Abstract: Clustering is a classic data structure analysis method in machine learning and pattern recognition. The anchor-based graph has improved graph-based clustering accuracy recently. The progressive self-supervised clustering method with novel category discovery (PSSCNCD) uses…
Abstract: A ParametRIc MAnifold Learning (PRIMAL) algorithm for Gaussian mixture models (GMM) assumes that GMMs lie on or near a manifold of probability distributions generated from a low-dimensional hierarchical latent space through parametric mappings. Latent…
Abstract: DCD is a popular optimization method. Sequential DCD makes parallelization difficult. Due to concurrent coordinate updates, running multiple DCD threads on batches of data elements causes inaccuracy and slow convergence. Some parallelization methods use…
Abstract: Attention helps model sequences and learn a deep representation. Recent studies show that attention values do not always match intuition in tasks like machine translation and sentiment classification. This study uses deep reinforcement learning…
Abstract: Multiview subspace clustering (MVSC) is a new method that finds the subspace in multiview data and clusters it. Many MVSC methods have been proposed recently, but most of them cannot explicitly preserve locality in…
Abstract: Unsupervised cross-domain fault diagnosis has been studied recently. It learns transferable features that reduce distribution inconsistency between source and target domains without target supervision. Most cross-domain fault diagnosis methods assume source and target fault…
Abstract: Deep image prior (DIP) uses a deep convolutional network (ConvNet) structure as an image prior. DIP empirically proves ConvNet structures work for image restoration. Why the DIP works well is unknown. Convolution’s benefit in…
Abstract: Unbalanced data can strongly bias classifiers toward the majority class. GAN methods have been proposed to address this issue. Existing GAN-based methods fail to exploit the classifier-generator relationship. This article proposes a novel three-player…
Abstract: Recently, AI has garnered attention. Privacy, security, and model fairness issues have arisen alongside its advances. Differential privacy, a promising mathematical model, can help solve these problems. For this reason, differential privacy has been…
Abstract: Because user preferences change over time, capturing them is essential to predicting user behavior. Many shallow and deep recommendation algorithms model user static and dynamic preferences separately, making it difficult to fuse them for…
WhatsApp us