A5079044133- Time Series Analysis and Forecasting
- Stock Market Forecasting Methods
- Neural Networks and Applications
- Human-Automation Interaction and Safety
- Forecasting Techniques and Applications
- Risk and Safety Analysis
- Anomaly Detection Techniques and Applications
- Perovskite Materials and Applications
- Advanced Chemical Sensor Technologies
- Nanowire Synthesis and Applications
- 2D Materials and Applications
- Image Processing and 3D Reconstruction
- Wireless Signal Modulation Classification
- Advanced Memory and Neural Computing
- HVDC Systems and Fault Protection
- Ga2O3 and related materials
- Network Security and Intrusion Detection
- Face and Expression Recognition
- Graphene research and applications
- Animal Nutrition and Physiology
- Infrared Target Detection Methodologies
- Complex Systems and Time Series Analysis
- Conducting polymers and applications
- Organic Electronics and Photovoltaics
- Remote Sensing and Land Use
University of Science and Technology of China
2022-2025
Chinese Academy of Sciences
2022-2025
China Agricultural University
2024
Beihang University
2023-2024
Nanjing University of Aeronautics and Astronautics
2019
North China Electric Power University
2019
This paper shows that time series forecasting Transformer (TSFT) suffers from severe over-fitting problem caused by improper initialization method of unknown decoder inputs, especially when handling non-stationary series. Based on this observation, we propose GBT, a novel two-stage framework with Good Beginning. It decouples the prediction process TSFT into two stages, including Auto-Regression stage and Self-Regression to tackle different statistical properties between input sequences....
Machine learning is the core of artificial intelligence. Using optical signals for training and converting them into electrical inference, combines strengths both, thus can greatly improve machine efficiency. Optoelectronic memories are hardware foundation this strategy. However, existing optoelectronic cannot modulate a large number non-volatile resistive states using ultra-short ultra-dim light pulses, leading to low accuracy, slow computing speed high energy consumption. Here, we...
Ambient solution-processed conductive materials with a sufficient low work function are essential to facilitate electron injection in electronic and optoelectronic devices but challenging. Here, we design an electrically conducting ambient-stable polymer electrolyte ultralow down 2.2 eV, which arises from heavy n-doping of dissolved salts matrix. Such can be solution processed into uniform smooth films on various conductors including graphene, metal oxides, polymers metals substantially...
With the development of Internet Things (IoT) systems, precise long-term forecasting method is requisite for decision makers to evaluate current statuses and formulate future policies. Currently, Transformer MLP are two paradigms deep time-series former one more prevailing in virtue its exquisite attention mechanism encoder-decoder architecture. However, data scientists seem be willing dive into research encoder, leaving decoder unconcerned. Some researchers even adopt linear projections...
In the human-machine system of armored vehicles, cognitive performance state crews is crucial for personnel security and combat efficiency. The purpose this research was to establish a real-time assessment performances vehicle crews, consisting data input module, processing visualization scheduling module. Forty subjects were recruited cooperate execute cross-platform strike task in virtual simulation platform. physiological operation behavior collected during experiment process. To realize...
How to handle time features shall be the core question of any series forecasting model. Ironically, it is often ignored or misunderstood by deep-learning based models, even those baselines which are state-of-the-art. This behavior makes their inefficient, untenable and unstable. In this paper, we rigorously analyze three prevalent but deficient/unfounded deep mechanisms methods from view properties, including normalization methods, multivariate input sequence length. Corresponding...
This paper presents FDNet: a Focal Decomposed Network for efficient, robust and practical time series forecasting. We break away from conventional deep forecasting formulas which obtain prediction results universal feature maps of input sequences. In contrary, FDNet neglects correlations elements only extracts fine-grained local features sequence. show that: (1) Deep with sequence is feasible upon theoretical basis. (2) By abandoning global coarse-grained maps, overcomes distribution shift...
The metal/germanium (Ge) photodetectors have attracted much attention for their potential applications in on-chip optoelectronics. One critical issue is the relatively large dark current due to limited Schottky barrier height of junction, which mainly caused by small bandgap Ge and Fermi energy level pinning effect between metal Ge. main technique solve this problem insert a thin interlayer However, so far, still when using bulk-material insertion layer, while two-dimensional area layer too...
Aiming at the problem that traditional classifier based on decision tree for radar jamming recognition needs prior information and manual intervention, a signal method fuzzy clustering is proposed in this paper. After establishing multidimensional sets of parameter features which are extracted from both time frequency domain, Fuzzy C-means (FCM) algorithm then employed building tree. Moreover, improved ID3 standard gain also used to complete The realizes an automatic design trees avoids...
This paper presents FDNet: a Focal Decomposed Network for efficient, robust and practical time series forecasting. We break away from conventional deep forecasting formulas which obtain prediction results universal feature maps of input sequences. In contrary, FDNet neglects correlations elements only extracts fine-grained local features sequence. show that: (1) Deep with sequence is feasible upon theoretical basis. (2) By abandoning global coarse-grained maps, overcomes distribution shift...
This paper shows that time series forecasting Transformer (TSFT) suffers from severe over-fitting problem caused by improper initialization method of unknown decoder inputs, esp. when handling non-stationary series. Based on this observation, we propose GBT, a novel two-stage framework with Good Beginning. It decouples the prediction process TSFT into two stages, including Auto-Regression stage and Self-Regression to tackle different statistical properties between input sequences.Prediction...
With the development of Internet Things (IoT) systems, precise long-term forecasting method is requisite for decision makers to evaluate current statuses and formulate future policies. Currently, Transformer MLP are two paradigms deep time-series former one more prevailing in virtue its exquisite attention mechanism encoder-decoder architecture. However, data scientists seem be willing dive into research encoder, leaving decoder unconcerned. Some researchers even adopt linear projections...