A5001979415- Recommender Systems and Techniques
- Topic Modeling
- Advanced Bandit Algorithms Research
- Machine Learning in Healthcare
- Image and Video Quality Assessment
- Human Mobility and Location-Based Analysis
- Innovative Teaching and Learning Methods
- Data Stream Mining Techniques
- Sentiment Analysis and Opinion Mining
- Advanced Graph Neural Networks
- Image Retrieval and Classification Techniques
- Consumer Market Behavior and Pricing
- Expert finding and Q&A systems
- Machine Learning and Data Classification
- Mobile Crowdsensing and Crowdsourcing
- Advanced Text Analysis Techniques
- Technology and Data Analysis
- Complex Network Analysis Techniques
- Intelligent Tutoring Systems and Adaptive Learning
- Video Analysis and Summarization
- Multimodal Machine Learning Applications
- Advanced Multi-Objective Optimization Algorithms
- Context-Aware Activity Recognition Systems
- Web Data Mining and Analysis
- Simulation Techniques and Applications
Jilin University
2024
City University of Hong Kong
2024
OriginWater (China)
2024
Kuaishou (China)
2024
University Town of Shenzhen
2024
Tsinghua University
2013-2024
Jilin Medical University
2024
University of Science and Technology Beijing
2013
Multi-domain recommendation and multi-task have demonstrated their effectiveness in leveraging common information from different domains objectives for comprehensive user modeling. Nonetheless, the practical usually faces multiple tasks simultaneously, which cannot be well-addressed by current methods. To this end, we introduce M3oE, an adaptive Multi-task Mixture-of-Experts framework. M3oE integrates multi-domain information, maps knowledge across tasks, optimizes objectives. We leverage...
Contrastive learning has recently emerged as an effective strategy for improving the performance of sequential recommendation. However, traditional models commonly construct contrastive loss by directly optimizing human-designed positive and negative samples, resulting in a model that is overly sensitive to heuristic rules. To address this limitation, we propose novel soft framework recommendation article. Our main idea extend point-wise contrast region-level comparison, where aim identify...
In Recommender System (RS) applications, reinforcement learning (RL) has recently emerged as a powerful tool, primarily due to its proficiency in optimizing long-term rewards.Nevertheless, it suffers from instability the process, stemming intricate interactions among bootstrapping, off-policy training, and function approximation.Moreover, multi-reward recommendation scenarios, designing proper reward setting that reconciles inner dynamics of various tasks is quite intricate.To this end, we...
In recent years, graph contrastive learning (GCL) has received increasing attention in recommender systems due to its effectiveness reducing bias caused by data sparsity. However, most existing GCL models rely on heuristic approaches and usually assume entity independence when constructing views. We argue that these methods struggle strike a balance between semantic invariance view hardness across the dynamic training process, both of which are critical factors learning. To address above...
In recommender systems, reinforcement learning solutions have shown promising results in optimizing the interaction sequence between users and system over long-term performance. For practical reasons, policy's actions are typically designed as recommending a list of items to handle users' frequent continuous browsing requests more efficiently. this list-wise recommendation scenario, user state is updated upon every request corresponding MDP formulation. However, request-level formulation...
In scenarios involving sequence recommendations on large screen devices, such as tablets or TVs, the equipment is often shared among multiple users. This sharing leads to a mixture of behaviors from different users, posing significant challenges recommendation systems, especially when clear supervisory signals for distinguishing users are absent. Current solutions tend either operate in an unsupervised manner rely constructed that not entirely reliable. Moreover, peculiarities short video...
Although numerous studies have been conducted on learning resources recommendation in E-Learning, research extending this investigation into usage of users' contextual information is rare. This paper presents an innovative architecture intelligent personalized context-aware system E-Learning environment. The offers users by recommending materials, tutors, or other learners with common interests combining social tags from external network and about materials autologous collaborative tagging...
Conversational recommender systems (CRSs) aim to recommend high-quality items users through a dialogue interface. It usually contains multiple sub-tasks, such as user preference elicitation, recommendation, explanation, and item information search. To develop effective CRSs, there are some challenges: 1) how properly manage sub-tasks; 2) effectively solve different 3) correctly generate responses that interact with users. Recently, Large Language Models (LLMs) have exhibited an unprecedented...
Location-Based service (LBS) provides the fundamental services for upcoming ubiquitous environments. However, most existing patterns location are driven and supported by a LBS provider only but lacking of user participation personalized recommendation. In this paper, we propose novel intelligent location-based to address these issues. Our solution is based on ontology which utilized model represent knowledge about user's individual preferences place-related information. The supports as...
In recommender systems, reinforcement learning solutions have shown promising results in optimizing the interaction sequence between users and system over long-term performance. For practical reasons, policy's actions are typically designed as recommending a list of items to handle users' frequent continuous browsing requests more efficiently. this list-wise recommendation scenario, user state is updated upon every request corresponding MDP formulation. However, request-level formulation...
In the landscape of Recommender System (RS) applications, reinforcement learning (RL) has recently emerged as a powerful tool, primarily due to its proficiency in optimizing long-term rewards. Nevertheless, it suffers from instability process, stemming intricate interactions among bootstrapping, off-policy training, and function approximation. Moreover, multi-reward recommendation scenarios, designing proper reward setting that reconciles inner dynamics various tasks is quite intricate....
ChatGPT has achieved remarkable success in natural language understanding. Considering that recommendation is indeed a conversation between users and the system with items as words, which similar underlying pattern ChatGPT, we design new chat framework item index level for task. Our novelty mainly contains three parts: model, training inference. For model part, adopt Generative Pre-training Transformer (GPT) sequential user modular to capture personalized information. two-stage paradigm of...
Recommender selects and presents top-K items to the user at each online request, a recommendation session consists of several sequential requests. Formulating as Markov decision process solving it by reinforcement learning (RL) framework has attracted increasing attention from both academic industry communities. In this paper, we propose RL-based industrial short-video recommender ranking framework, which models maximizes watch-time in an environment multi-aspect preferences collaborative...
We propose a general model-agnostic Contrastive learning framework with Counterfactual Samples Synthesizing (CCSS) for modeling the monotonicity between neural network output and numerical features which is critical interpretability effectiveness of recommender systems. CCSS models via two-stage process: synthesizing counterfactual samples contrasting samples. The two techniques are naturally integrated into framework, forming an end-to-end training process. Abundant empirical tests...
Recommender systems filter out information that meets user interests. However, users may be tired of the recommendations are too similar to content they have been exposed in a short historical period, which is so-called fatigue. Despite significance for better experience, fatigue seldom explored by existing recommenders. In fact, there three main challenges addressed modeling fatigue, including what features support it, how it influences interests, and its explicit signals obtained. this...
Recommender systems aim to fulfill the user's daily demands. While most existing research focuses on maximizing engagement with system, it has recently been pointed out that how frequently users come back for service also reflects quality and stability of recommendations. However, optimizing this user retention behavior is non-trivial poses several challenges including intractable leave-and-return activities, sparse delayed signal, uncertain relations between users' their immediate feedback...