A5027724935- Reinforcement Learning in Robotics
- Energy Load and Power Forecasting
- Evolutionary Algorithms and Applications
- Adaptive Dynamic Programming Control
- Model Reduction and Neural Networks
- Advanced Bandit Algorithms Research
- Adversarial Robustness in Machine Learning
- Anomaly Detection Techniques and Applications
- Smart Grid and Power Systems
- Image and Video Stabilization
- Educational Technology and Assessment
- Video Surveillance and Tracking Methods
- Wind and Air Flow Studies
- Computational Physics and Python Applications
- Explainable Artificial Intelligence (XAI)
- Autonomous Vehicle Technology and Safety
- Wind Energy Research and Development
- Industrial Technology and Control Systems
- Optimization and Search Problems
- RFID technology advancements
- Electric Power System Optimization
- Scheduling and Timetabling Solutions
- Radiation Effects in Electronics
- Domain Adaptation and Few-Shot Learning
- Generative Adversarial Networks and Image Synthesis
Harbin Engineering University
2023
Harbin Institute of Technology
2019-2022
Nanjing University of Information Science and Technology
2021-2022
George Washington University
2022
Nanjing University of Science and Technology
2006
Wind speed prediction is of great importance because it affects the efficiency and stability power systems with a high proportion wind power. Temporal-spatial features contain rich information; however, their use to predict remains one most challenging less studied areas. This paper investigates problem predicting speeds for multiple sites using temporal spatial proposes novel two-layer attention-based long short-term memory (LSTM), termed 2Attn-LSTM, unified framework encoder decoder...
Safety helmet-wearing detection is an essential part of the intelligent monitoring system. To improve speed and accuracy detection, especially small targets occluded objects, it presents a novel efficient detector model. The underlying core algorithm this model adopts YOLOv5 (You Only Look Once version 5) network with best comprehensive performance. It improved by adding attention mechanism, CIoU (Complete Intersection Over Union) Loss function, Mish activation function. First, applies...
Efficient exploration remains a challenging problem in reinforcement learning, especially for tasks where extrinsic rewards from environments are sparse or even totally disregarded. Significant advances based on intrinsic motivation show promising results simple but often get stuck with multimodal and stochastic dynamics. In this work, we propose variational dynamic model the conditional inference to multimodality stochasticity. We consider environmental state-action transition as generative...
With the advent of Deep Neural Network (DNN) accelerators, permanent faults are increasingly becoming a serious challenge for DNN hardware accelerator, as they can severely degrade inference accuracy. The State-of-the-art works address this issue by adding homogeneous redundant Processing Elements (PEs) to accelerator's central computing array, or bypassing faulty PEs directly. However, such designs induce loss, extra cost, and performance overhead. Moreover, current able only deal with...
Improving short-term wind speed prediction accuracy and stability remains a challenge for forecasting researchers. This paper proposes new variational mode decomposition (VMD)-attention-based spatio-temporal network (VASTN) method that takes advantage of both temporal spatial correlations speed. First, VASTN is hybrid model combines VMD, squeeze-and-excitation (SENet), attention mechanism (AM)-based bidirectional long memory (BiLSTM). initially employs VMD to decompose the matrix into series...
Unmanned Combat Aerial Vehicle (UCAV) dogfight, which refers to a fight between two or more UCAVs usually at close quarters, plays decisive role on the aerial battlefields. With evolution of artificial intelligence, dogfight progressively transits towards intelligent and autonomous modes. However, development policy learning is hindered by challenges such as weak exploration capabilities, low efficiency, unrealistic simulated environments. To overcome these challenges, this paper proposes...
Anderson mixing has been heuristically applied to reinforcement learning (RL) algorithms for accelerating convergence and improving the sampling efficiency of deep RL. Despite its heuristic improvement convergence, a rigorous mathematical justification benefits in RL not yet put forward. In this paper, we provide deeper insights into class acceleration schemes built on that improve algorithms. Our main results establish connection between quasi-Newton methods prove increases radius policy...
This paper proposes a spatio-temporal model (VCGA) based on variational mode decomposition (VMD) and attention mechanism. The proposed prediction combines squeeze-and-excitation network to extract spatial features gated recurrent unit capture temporal dependencies. Primarily, the VMD can reduce instability of original wind speed data mechanism functions strengthen impact important information. In addition, act avoid decline in accuracy. Finally, VCGA trains result derives final results after...