A5086273618- Recommender Systems and Techniques
- Caching and Content Delivery
- Domain Adaptation and Few-Shot Learning
- Data Visualization and Analytics
- Advanced Bandit Algorithms Research
- Mental Health Research Topics
- Machine Learning and ELM
- Consumer Market Behavior and Pricing
- IoT and Edge/Fog Computing
- 3D IC and TSV technologies
- Multimodal Machine Learning Applications
- Auction Theory and Applications
- Advanced Graph Neural Networks
- VLSI and FPGA Design Techniques
- Computer Graphics and Visualization Techniques
- Advancements in Photolithography Techniques
- Reinforcement Learning in Robotics
- Sharing Economy and Platforms
- Customer churn and segmentation
- Machine Learning and Algorithms
- Music and Audio Processing
- Adversarial Robustness in Machine Learning
- Generative Adversarial Networks and Image Synthesis
- Stock Market Forecasting Methods
- Sparse and Compressive Sensing Techniques
Tsinghua University
2021-2025
Tsinghua–Berkeley Shenzhen Institute
2021-2025
University of Southern California
2024
The paper proposes a physics-informed model to predict battery lifetime trajectories by computing thermodynamic and kinetic parameters, saving costly data that has not been established for sustainable manufacturing, reuse, recycling.
This study proposes a novel federated learning framework for optimizing intelligent edge computing resource scheduling. The addresses the challenges of device heterogeneity, non-IID data distribution, and communication overhead in environments. We introduce an adaptive client selection mechanism considering computational capabilities, energy status, quality. A personalized model training approach is implemented to handle effectively using multi-task local batch normalization layers....
This study discussion point of this paper is to make an in-depth analysis the development impact Internet Things combined with edge computing and artificial intelligence. In process, importance criticality data processing decision making as well challenges faced should be elaborated respectively. With rapid popularization devices, has brought more innovative solutions for different application scenarios such intelligent furniture industrialization, automatic driving transportation by...
This paper discusses the application and advantages of machine learning in anomaly detection network security traffic. By summarizing existing methods techniques detection, this focuses on progress clustering, classification, statistics, information theory research. In particular, innovations data preprocessing, feature selection, algorithm design, such as experimental validation based an improved KNN algorithm, demonstrate potential improving accuracy efficiency. future, amount increases...
Adapting deep neural networks to the changing environments is critical in practical utility, especially for online web applications, where data distribution changes gradually due evolving environments. For instance, photos of cellphones change over years appearance changes. This paper deals with such a problem via active gradual domain adaptation, learner continually and actively selects most informative labels from target enhance labeling efficiency utilizes both labeled unlabeled samples...
With the gradual development and integration of artificial intelligence into various industries, there is also a great range in financial industry. Therefore, this article focuses on trend prediction model risk management problems deep reinforcement learning (DRL), one largest branches intelligence, cryptocurrency market. In addition, experimental part paper, machine Long short-term memory network (LSTM) used to make effective time series analysis relevant data market, so as large-scale...
Despite the broad interest in meta-learning, generalization problem remains one of significant challenges this field. Existing works focus on meta-generalization to unseen tasks at meta-level by regularizing meta-loss, while ignoring that adapted models may not generalize task domains adaptation level. In paper, we propose a new regularization mechanism for meta-learning - Minimax-Meta Regularization, which employs inverted inner loop and ordinary outer during training. particular, makes...
This paper presents a new method for chip floorplanning optimization using deep learning (DRL) combined with graph neural networks (GNNs). The plan addresses the challenges of traditional floor plans by applying AI to space design and intelligent decisions. Three-head network architecture, including policy network, cost reconstruction head, is introduced improve feature extraction overall performance. GNNs are employed state representation extraction, enabling capture intricate topological...
This study discussion point of this paper is to make an in-depth analysis the development impact Internet Things combined with edge computing and artificial intelligence. In process, importance criticality data processing decision making as well challenges faced should be elaborated respectively. With rapid popularization devices, has brought more innovative solutions for different application scenarios such intelligent furniture industrialization, automatic driving transportation by...
Graph neural networks (GNNs) have achieved tremendous success in the task of graph classification and its diverse downstream real-world applications. Despite huge learning representations, current GNN models demonstrated their vulnerability to potentially existent adversarial examples on graph-structured data. Existing approaches are either limited structure attacks or restricted local informatio, urging for design a more general attack framework classification, which faces significant...
The rapid growth of rich multimedia data in today's Internet, especially video traffic, has challenged the content delivery networks (CDNs). Caching serves as an important means to reduce user access latency so enable faster downloads. Motivated by dynamic nature real-world edge traces, this paper introduces a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">provably well</i> online caching policy environments where: 1) popularity is highly...
We study the budget allocation problem in online marketing campaigns that utilize previously collected offline data. first discuss long-term effect of optimizing decisions setting. To overcome challenge, we propose a novel game-theoretic value-based reinforcement learning method using mixed policies. The proposed reduces need to store infinitely many policies previous methods only constantly policies, which achieves nearly optimal policy efficiency, making it practical and favorable for...
The application of artificial intelligence (AI) continues to expand across various industries, especially in enhancing user experience and optimizing business processes. Through deep learning machine algorithms, companies are able analyze behavior data provide personalized recommendations, which effectively improve customer satisfaction loyalty. This data-driven approach enables businesses stand out a highly competitive market. paper explores the key role learning-based recommendation...
This paper presents a new method for chip floorplanning optimization using deep learning (DRL) combined with graph neural networks (GNNs). The plan addresses the challenges of traditional floor plans by applying AI to space design and intelligent decisions. Three-head network architecture, including policy network, cost reconstruction head, is introduced improve feature extraction overall performance. GNNs are employed state representation extraction, enabling capture intricate topological...
The content request patterns perceived by edge devices are becoming highly dynamic, especially for emerging short video platforms compared to traditional platforms. This calls caching policies that can continuously adapt dynamic environments, challenging previously popular reinforcement learning (RL)-based policies. A straightforward solution, i.e., repeatedly restarting and training RL agents, would fail converge timely while meeting the observed adaptation process. Offering transferable...
Models trained with offline data often suffer from continual distribution shifts and expensive labeling in changing environments. This calls for a new online learning paradigm where the learner can continually adapt to environments limited labels. In this paper, we propose setting -- Online Active Continual Adaptation, aims distributions using both unlabeled samples active queries of To end, Self-Adaptive Mirror Descent (OSAMD), which adopts an teacher-student structure enable self-training...
Model-free RL-based recommender systems have recently received increasing research attention due to their capability handle partial feedback and long-term rewards. However, most existing has ignored a critical feature in systems: one user's on the same item at different times is random. The stochastic rewards property essentially differs from that classic RL scenarios with deterministic rewards, which makes much more challenging. In this paper, we first demonstrate simulator environment...
Current learning-based edge caching schemes usually suffer from dynamic content popularity, e.g., in the emerging short video platforms, users' request patterns shift significantly over time and across different edges. An intuitive solution for a specific local cache is to collect more histories other caches. However, uniformly merging these may not perform satisfactorily due heterogeneous distributions on To solve this problem, we propose collaborative framework. First, design...
Current learning-based edge caching schemes usually suffer from dynamic content popularity, e.g., in the emerging short video platforms, users' request patterns shift significantly over time and across different edges. An intuitive solution for a specific local cache is to collect more histories other caches. However, uniformly merging these may not perform satisfactorily due heterogeneous distributions on To solve this problem, we propose collaborative framework. First, design...
We study the budget allocation problem in online marketing campaigns that utilize previously collected offline data. first discuss long-term effect of optimizing decisions setting. To overcome challenge, we propose a novel game-theoretic value-based reinforcement learning method using mixed policies. The proposed reduces need to store infinitely many policies previous methods only constantly policies, which achieves nearly optimal policy efficiency, making it practical and favorable for...