A5100391202- Cosmology and Gravitation Theories
- Pulsars and Gravitational Waves Research
- Age of Information Optimization
- Quantum Computing Algorithms and Architecture
- Engineering Education and Technology
- IoT and Edge/Fog Computing
- Trade Secret Protection Methods
- Intellectual Property and Patents
- Psychology of Moral and Emotional Judgment
- Peer-to-Peer Network Technologies
- Seismic Waves and Analysis
- Law, AI, and Intellectual Property
- Artificial Intelligence in Healthcare
- Neural dynamics and brain function
- Ethics in Business and Education
- Human Pose and Action Recognition
- Distributed and Parallel Computing Systems
- Geophysics and Gravity Measurements
- Energy Efficient Wireless Sensor Networks
- Conflict Management and Negotiation
- Functional Brain Connectivity Studies
- Reinforcement Learning in Robotics
- Anomaly Detection Techniques and Applications
- Mobile Agent-Based Network Management
- Service-Oriented Architecture and Web Services
Sun Yat-sen University
2021-2024
Chongqing University of Posts and Telecommunications
2023
Jishou University
2011
Southerners on New Ground
2010
By collecting and processing smart healthcare data in real-time, mobile edge networks can bridge the gap between doctors patients, which is a representative consumer-level interaction scenario Industry 5.0. To improve flexibility timeliness of suchlike 5.0 services, microservices are deployed extensively on servers. The major challenge facing microservice deployment how to reduce cost. However, cost quality service (QoS) pair contradictory objectives, different scenarios have preferences for...
Prioritized Experience Replay (ER) has been empirically shown to improve sample efficiency across many domains and attracted great attention; however, there is little theoretical understanding of why such prioritized sampling helps its limitations. In this work, we take a deep look at the ER. supervised learning setting, show equivalence between error-based method for mean squared error uniform cubic power loss. We then provide insight into it improves convergence rate upon during early...
This paper tackles the problem of how to pre-train a model and make it generally reusable backbones for downstream task learning. In pre-training, we propose method that builds an agent-environment interaction by learning domain invariant successor features from agent's vast experiences covering various tasks, then discretize them into behavior prototypes which result in embodied set structure. To learning, (1) feature projection retains previous knowledge projecting new task's...