A5064223000- Recommender Systems and Techniques
- Advanced Bandit Algorithms Research
- Text and Document Classification Technologies
- Data Stream Mining Techniques
- Music and Audio Processing
- Advanced Graph Neural Networks
- Ovarian cancer diagnosis and treatment
- Machine Learning in Healthcare
- Topic Modeling
- Multimodal Machine Learning Applications
- Endometrial and Cervical Cancer Treatments
- Caching and Content Delivery
- Image Retrieval and Classification Techniques
- Cervical Cancer and HPV Research
Hefei University of Technology
2023-2025
Institute of Art
2024
China Academy of Information and Communications Technology
2024
Tabular data is one of the most widely used formats across various domains such as bioinformatics, healthcare, and marketing. As artificial intelligence moves towards a data-centric perspective, improving quality essential for enhancing model performance in tabular data-driven applications. This survey focuses on optimization, specifically exploring reinforcement learning (RL) generative approaches feature selection generation fundamental techniques refining spaces. Feature aims to identify...
Collaborative Filtering~(CF) typically suffers from the significant challenge of popularity bias due to uneven distribution items in real-world datasets. This leads a accuracy gap between popular and unpopular items. It not only hinders accurate user preference understanding but also exacerbates Matthew effect recommendation systems. To alleviate bias, existing efforts focus on emphasizing or separating correlation item representations their popularity. Despite effectiveness, works still...
As its availability and generality in online services, implicit feedback is more commonly used recommender systems. However, usually presents noisy samples real-world recommendation scenarios (such as misclicks or non-preferential behaviors), which will affect precise user preference learning. To overcome the problem, a popular solution based on dropping model training phase, follows observation that have higher losses than clean samples. Despite effectiveness, we argue this still has...
Multimedia-based recommendation provides personalized item suggestions by learning the content preferences of users.With proliferation digital devices and APPs, a huge number new items are created rapidly over time.How to quickly provide recommendations for at inference time is challenging.What's worse, real-world exhibit varying degrees modality missing (e.g., many short videos uploaded without text descriptions).Though efforts have been devoted multimedia-based recommendations, they either...
Recommender systems face a daunting challenge when entities (users or items) without any historical interactions, known as the " <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Completely Cold-Start Problem</i> ". Due to absence of collaborative signals, Collaborative Filtering (CF) schema fails deduce user preferences item characteristics for such cold entities. A common solution is incorporating auxiliary discrete attributes bridge spread...
Multimedia-based recommendation models learn user and item preference representation by fusing both the user-item collaborative signals multimedia content signals. In real scenarios, cold items appear in test stage without any interaction record. How to perform is challenging as training have different data distributions. These hybrid representations contained auxiliary signals, so current solutions designed alignment functions transfer learned items. Despite effectiveness, we argue that...
Collaborative Filtering (CF) plays a crucial role in modern recommender systems, leveraging historical user-item interactions to provide personalized suggestions. However, CF-based methods often encounter biases due imbalances training data. This phenomenon makes tend prioritize recommending popular items and performing unsatisfactorily on inactive users. Existing works address this issue by rebalancing samples, reranking recommendation results, or making the modeling process robust bias....
Multimedia recommendation, which incorporates various modalities (e.g., images, texts, etc.) into user or item representation to improve recommendation quality, has received widespread attention. Recent methods mainly focus on cross-modal alignment with self-supervised learning obtain higher quality representation. Despite remarkable performance, we argue that there is still a limitation: completely aligning undermines modality-unique information. We consider right, but it should not be the...