Projects

Abstract: The discriminative mixture variational autoencoder (DMVAE) is a deep probability model for semisupervised learning feature extraction. Encoding, decoding, and classification modules comprise the DMVAE. The encoder projects the observation to the latent space in…

Abstract: Spatial-temporal data is increasingly available due to the rapid development of GPS, mobile devices, and remote sensing. Human mobility understanding, smart transportation, urban planning, public safety, health care, and environmental management depend on spatio-temporal…

Abstract: Multimedia, machine learning, and data mining communities are focusing on multi-view clustering. Multi-view subspace clustering (MVSC) is a popular multi-view clustering algorithm because it can reveal the intrinsic low-dimensional clustering structure hidden across views….

Abstract: This article presents a data-driven method for computing reachable sets where active learning (AL) is used to reduce the computational burden. Set-based methods used to estimate reachable sets typically do not scale well with…

Abstract: Navigation apps use transportation recommendation. Previous transportation recommendation solutions only consider routes in one transportation mode (uni-modal, e.g., taxi, bus, cycle) and ignore situational context, resulting in poor user experience. We propose $\mathsf {Hydra}$,…

Abstract: Due to the low signal-to-noise ratio of scalp EEG, brain–computer interface (BCI) development requires accurate electroencephalogram (EEG) pattern decoding for specific mental tasks. Machine learning can optimize EEG patterns for better decoding. Existing algorithms…

Abstract: Distributing large sparse matrix operations in scientific computing is like hypergraph partitioning. Hypergraphs have “hyperedges” that connect any number of nodes. Thus, solving or approximating hypergraph partitioning is NP-hard. Current algorithms solve this problem…

Abstract: Daily communication is influenced by many factors. These factors allow conversational systems to generate positive responses and build friendly relationships with users. This study examines emotion and intention as response generators. We create a…

Abstract: Given a user’s purchase/rating trajectories, sequential recommendation recommends the next few items they’re most likely to buy/review. It helps users choose their favorites. This manuscript uses hybrid associations models (HAM) to generate sequential recommendations…

Abstract: Entity resolution (ER), a classification problem, is difficult on real data due to dirty values. The best ER solutions use learning models like deep neural networks, which need lots of labeled training data. High-quality…

Abstract: Machine learning libraries need static and dynamic clustering algorithms. The sliding window model or simpler models have dominated dynamic machine learning and data mining algorithm development. Many real-world applications require arbitrary deletions and insertions….

Abstract: Factorization Machines (FMs) help recommender systems overcome cold-start and data sparsity by incorporating side information. Traditional FMs use the inner product to model second-order interactions between feature vectors. Inner products violate feature vector triangle…