A5101968769- Neural Networks and Applications
- GaN-based semiconductor devices and materials
- Evolutionary Algorithms and Applications
- Semiconductor materials and devices
- Metaheuristic Optimization Algorithms Research
- Electrohydrodynamics and Fluid Dynamics
- Plasma Diagnostics and Applications
- Advanced Neural Network Applications
- Ga2O3 and related materials
- Dust and Plasma Wave Phenomena
- Anomaly Detection Techniques and Applications
- Natural Language Processing Techniques
- Silicon Carbide Semiconductor Technologies
- Topic Modeling
- Machine Learning in Materials Science
- Nanowire Synthesis and Applications
- Music and Audio Processing
- Advancements in Semiconductor Devices and Circuit Design
- Neural Networks and Reservoir Computing
- Insect and Arachnid Ecology and Behavior
- Industrial Vision Systems and Defect Detection
- Mathematics, Computing, and Information Processing
- Machine Learning and Data Classification
- Statistics Education and Methodologies
- Analog and Mixed-Signal Circuit Design
University of Chinese Academy of Sciences
2018-2025
Institute of Semiconductors
2017-2025
Chinese Academy of Sciences
2017-2025
Hebei University of Engineering
2018-2023
Handan College
2017
University of Freiburg
2015
Dalian University
2009
Dalian University of Technology
2009
This study focuses on the InAlN/GaN/AlGaN heterostructures, initiating from polarization energy band tailoring engineering to establish a theoretical model. An AlxGa1-xN back-barrier is introduced augment confinement ability of two-dimensional electron gas (2DEG) in conduction channel, enhance power output, and reduce leakage buffer layer. The primary investigation explores impact GaN channel layer’s thickness Al content x layer properties sheet density 2DEG. It also examines formation...
Finding a concise and interpretable mathematical formula that accurately describes the relationship between each variable predicted value in data is crucial task scientific research, as well significant challenge artificial intelligence. This problem referred to symbolic regression, which an NP-hard problem. In previous year, novel regression methodology utilizing Monte Carlo Tree Search (MCTS) was advanced, achieving state-of-the-art results on diverse range of datasets. although this...
Gallium nitride- (GaN) based high electron mobility transistors (HEMTs) provide a good platform for biological detection. In this work, both Au-gated AlInN/GaN HEMT and AlGaN/GaN biosensors are fabricated the detection of deoxyribonucleic acid (DNA) hybridization. The biosensor exhibits higher sensitivity in comparison with biosensor. For former, drain-source current ( V) shows clear decrease 69 μA upon introduction 1 μmolL (μM) complimentary DNA to probe at sensor area, while latter it is...
The temperature uniformity and heating efficiency in a high-temperature epitaxial growth system were investigated by modeling simulating. finite element method (FEM) was used to calculate the distribution of magnetic field reactor system. simulation results showed that due skin effect heat conduction conventional susceptor, wafer on susceptor is not uniform. However, can be greatly improved adding an air gap below wafer. existence effectively reduced center wafer, its radius studied. By...
Artificial neural networks (ANNs) have permeated various disciplinary domains, ranging from bioinformatics to financial analytics, where their application has become an indispensable facet of contemporary scientific research endeavors. However, the inherent limitations traditional arise due relatively fixed network structures and activation functions. 1, The type function is single fixed, which leads poor "unit representation ability" network, it often used solve simple problems with very...
Download This Paper Open PDF in Browser Add to My Library Share: Permalink Using these links will ensure access this page indefinitely Copy URL DOI
Symbolic regression aims to derive interpretable symbolic expressions from data in order better understand and interpret data. %which plays an important role knowledge discovery machine learning. In this study, a network called PruneSymNet is proposed for regression. This novel neural whose activation function consists of common elementary functions operators. The whole differentiable can be trained by gradient descent method. Each subnetwork the corresponds expression, our goal extract such...
Mathematical formulas are the crystallization of human wisdom in exploring laws nature for thousands years. Describing complex with a concise mathematical formula is constant pursuit scientists and great challenge artificial intelligence. This field called symbolic regression. Symbolic regression was originally formulated as combinatorial optimization problem, GP reinforcement learning algorithms were used to solve it. However, sensitive hyperparameters, these two types inefficient. To this...
The mathematical formula is the human language to describe nature and essence of scientific research. Finding formulas from observational data a major demand research challenge artificial intelligence. This area called symbolic regression. Originally regression was often formulated as combinatorial optimization problem solved using GP or reinforcement learning algorithms. These two kinds algorithms have strong noise robustness ability good Versatility. However, inference time usually takes...
Download This Paper Open PDF in Browser Add to My Library Share: Permalink Using these links will ensure access this page indefinitely Copy URL DOI
Noise ubiquitously exists in signals due to numerous factors including physical, electronic, and environmental effects. Traditional methods of symbolic regression, such as genetic programming or deep learning models, aim find the most fitting expressions for these signals. However, often overlook noise present real-world data, leading reduced accuracy. To tackle this issue, we propose \textit{\textbf{D}eep Symbolic Regression against \textbf{N}oise via \textbf{C}ontrastive \textbf{L}earning...
Mathematical formulas serve as the means of communication between humans and nature, encapsulating operational laws governing natural phenomena. The concise formulation these is a crucial objective in scientific research an important challenge for artificial intelligence (AI). While traditional neural networks (MLP) excel at data fitting, they often yield uninterpretable black box results that hinder our understanding relationship variables x predicted values y. Moreover, fixed network...
Large Language Models (LLMs) have made incredible strides recently in understanding and reacting to user intents. However, these models typically excel English not been specifically trained for medical applications, leading suboptimal performance responding inquiries such as diagnostic queries drug recommendations. In this paper, we propose DoctorGPT, a domain-specific large language model tailored question-answering tasks. DoctorGPT leverages the open-source Baichuan2 its foundational...
This work reports a preliminary investigation of energy bands AlxGal-xN/GaN heterojunction based on the use artificial neural networks (ANN). Numerical simulations were used to generate training and testing dataset for ANN model. The input parameters are Al content, thicknesses AlxGal-xN barrier layer position description two-Layer Materials, respectively. outputs conduction band profile AlxGal_xN/GaN channel Two-dimensional electron gas (2DEG) concentration distributions. results show that...
A self-consistent two-dimensional (2D) collisionless fluid model is developed to simulate the characteristics of a dual frequency capacitive sheath over an electrode with cylindrical hole. The consists 2D time-dependent equations coupled Poisson's equation, in which low-frequency (LF) and high-frequency current sources are applied electrode. Thus, so-called equivalent circuit coupling will be able self-consistently determine relationship between instantaneous voltage on powered thickness....
A self-consistent two-dimensional (2D) collisionless fluid model is developed to simulate the effects of low-frequency (LF) power on a dual frequency (DF) capacitive sheath over an electrode with cylindrical hole. In this paper, time-averaged potential, electric field (E-field), ion density in sheath, and energy distributions (IEDs) at center hole's bottom are calculated compared for different LF powers. The results show that crucial determining structure. As decreases, potential drop...