- Recommender Systems and Techniques
- Advanced Graph Neural Networks
- Topic Modeling
- Semiconductor materials and devices
- Advanced battery technologies research
- Electrocatalysts for Energy Conversion
- Fuel Cells and Related Materials
- Mass Spectrometry Techniques and Applications
- Advanced Bandit Algorithms Research
- Electronic and Structural Properties of Oxides
- Ferroelectric and Negative Capacitance Devices
- Plasma Diagnostics and Applications
- Copper Interconnects and Reliability
- AI-based Problem Solving and Planning
- Block Copolymer Self-Assembly
- Distributed Control Multi-Agent Systems
- Structural Engineering and Vibration Analysis
- Advanced Manufacturing and Logistics Optimization
- Membrane-based Ion Separation Techniques
- Structural Analysis and Optimization
- Scientific Computing and Data Management
- Artificial Intelligence in Healthcare and Education
- Multimodal Machine Learning Applications
- Mental Health via Writing
- Image and Video Quality Assessment
Chang'an University
2021-2025
Shijiazhuang Tiedao University
2025
Huazhong University of Science and Technology
2020-2024
Tianjin University
2012-2022
Southeast University
2022
University of Science and Technology of China
2021
The University of Texas at Austin
2016-2019
Lam Research (United States)
2019
Abstract Design and fabrication of hierarchically structured membranes with high proton conductivity is crucial to many energy‐relevant applications including exchange membrane fuel cell (PEMFC). Here, a series imidazole microcapsules (IMCs) tunable group loading, shell thickness, lumen size are synthesized incorporated into sulfonated poly(ether ether ketone) (SPEEK) matrix prepare composite membranes. The IMCs play two roles: i) Improving water retention properties the membrane. IMCs,...
Some recent news recommendation (NR) methods introduce a Pre-trained Language Model (PLM) to encode representation by following the vanilla pre-train and fine-tune paradigm with carefully-designed recommendation-specific neural networks objective functions. Due inconsistent task that of PLM, we argue their modeling has not well exploited abundant semantic information linguistic knowledge embedded in pre-training process. Recently, pre-train, prompt, predict paradigm, called prompt learning,...
Proton exchange membranes (PEM) with affordable and controllable proton conductivity under low humidity are crucial to the commercial application of PEM fuel cells. In this study, double-shelled polymer microcapsules bearing a carboxylic acid inner shell imidazole outer (PMC-Ns) synthesized via distillation–precipitation polymerization then incorporated into sulfonated poly(ether ether ketone) matrix fabricate composite membranes. The renders absorbed water lower chemical potential higher...
H2O2 and polarity are quite important in many physiological pathological processes, their relationship is complicated obscure for researchers. Thus, it vital challenging to achieve simultaneous detection of vivo. Herein, the first naphthalimide–triphenylamine-based dual-site fluorescent probe NATPA developed simultaneously imaging intracellular fluctuations. It exhibits excellent sensitivity (LOD = 44 nM), selectivity, fast response (15 min) a superior capacity detecting upon intramolecular...
Incorporating knowledge graph as side information has become a new trend in recommendation systems. Recent studies regard items entities of and leverage neural networks to assist item encoding, yet by considering each relation type individually. However, types are often too many sometimes one involves few entities. We argue that it is not efficient nor effective use every for encoding. In this paper, we propose VRKG4Rec model (Virtual Relational Knowledge Graphs Recommendation), which...
Abstract In order to study the fatigue performance of segmental precast concrete T‐section Beams, six test beams were designed and manufactured with construction technology (integral cast‐in‐place precast) key teeth type (large key‐joint double key‐joint) as control factors. Through two‐point loading test, ultimate bearing capacity beam is obtained, which provides a basis for upper limit lower load. After that, loads carried out on three in same batch static load compressive strain, steel...
This work targets the area selective atomic layer deposition (AS-ALD) of TiN onto HfO2 for use as word line in a memory device. Unlike other patterning processes, AS-ALD eliminates etching steps and also allows growth patterned films with precise thickness control. study investigates how differs on planar nonplanar surfaces. Using combination X-ray photoelectron spectroscopy, scanning electron microscopy, transmission we demonstrate way to confer selectivity substrate using surface features....
The authors report the deposition of 4.5-nm-thick cobalt (II) oxide on SiO2/Si(001) and MgO(001) substrates at 180–270 °C by atomic layer using bis(N-tert-butyl-N′-ethylpropionamidinato) water as coreactants. resulting CoO film is smooth carbon-free. can be reduced to Co metal hydrogen or deuterium gas 400–500 in a vacuum furnace, but high temperature processing causes dewetting, leading discontinuous islands rather than continuous films. Two low (∼200 °C) reduction methods are reported:...
The authors report the area-selective deposition of cobalt (II) oxide on polystyrene-patterned SiO2/Si and MgO(001) substrates at 180 °C by atomic layer (ALD) using bis(N-tert butyl, N′-ethylpropionamidinato) water as coreactants. patterned CoO films are carbon-free, smooth, were reduced with deuterium 220 to produce Co metal patterns without shape deformation. ALD is facile starting surfaces that features hydroxyl groups favoring nucleation growth. Polystyrene (PS) very effective in...
Abstract In this study, we evaluate the programming capabilities of OpenAI's GPT‐3.5 and GPT‐4 models using Swift‐based exam questions from a third‐year university course. The results indicate that both GPT generally outperform average student score, yet they do not consistently exceed performance top students. This comparison highlights areas where excel fall short, providing nuanced view their current proficiency. study also reveals surprising instances outperforms GPT‐4, suggesting...
Session-based recommendation (SBR) is to predict the next item for an anonymous sequence. Although many neural models have proven effectiveness in SBR task, how learn better items' embeddings still remains a key challenge due anonymity of sessions and sparsity users' behaviors. This paper proposes graph-based model, called Graph N eighborhood Routing Random Walk (GNRRW), which learns two kinds task. We first construct graph based on co-occurrences all sessions, we local embedding global each...
The design and optimization of highly nonlinear complex processes like plasma etching is challenging timeconsuming. Significant effort has been devoted to creating profile simulators facilitate the development etch recipes. Nevertheless, these are often difficult use in practice due large number unknown parameters discharge surface kinetics material, dependency rate on evolving front profile, disparate length scales system. Here, we expand a previously published, data informed, Bayesian...
A fast and inexpensive scheme for etch rate prediction using flexible continuum models Bayesian statistics is demonstrated. Bulk rates of MgO are predicted a steady-state model with volume-averaged plasma parameters classical Langmuir surface kinetics. Plasma particle kinetics modeled within global framework single component Metropolis Hastings methods limited data. The accuracy these predictions evaluated synthetic experimental data magnesium oxide in an ICP-RIE system. This approach...
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
In the burgeoning field of artificial intelligence (AI), understanding capabilities and limitations programming-oriented models is crucial. This paper presents a novel evaluation programming proficiency Generative Pretrained Transformer (GPT) models, specifically GPT-3.5 GPT-4, against coding problems varying difficulty levels drawn from Codewars. The experiments reveal distinct boundary at 3kyu level, beyond which these GPT struggle to provide solutions. These findings led proposal measure...
Session-based recommendation (SBR) aims to predict the next item based on user behaviors within a short time period. Most of existing solutions exploit relations among items and ignore attributes information (e.g. category). We observe that interests in may change frequently one session, but less so categories, argue introducing category into SBR alleviate data sparsity problem promote task next-item prediction. In this paper, we propose novel method called CDT-GNN, which incorporates...