A5100734874- Neural Networks and Applications
- Advanced Neural Network Applications
- Anomaly Detection Techniques and Applications
- Crystallization and Solubility Studies
- X-ray Diffraction in Crystallography
- Face and Expression Recognition
- Catalytic Processes in Materials Science
- Adversarial Robustness in Machine Learning
- Gaussian Processes and Bayesian Inference
- Machine Learning and Data Classification
- Catalysis and Oxidation Reactions
- Machine Learning and ELM
- Machine Learning in Healthcare
- Metaheuristic Optimization Algorithms Research
- Catalysis and Hydrodesulfurization Studies
- Supply Chain and Inventory Management
- Domain Adaptation and Few-Shot Learning
- Recommender Systems and Techniques
- Generative Adversarial Networks and Image Synthesis
- Artificial Intelligence in Games
- Groundwater flow and contamination studies
- Explainable Artificial Intelligence (XAI)
- Computational Drug Discovery Methods
- Wireless Communication Security Techniques
- Imbalanced Data Classification Techniques
Shenzhen Polytechnic
2024
Guizhou University
2021-2022
University of Manchester
2019-2021
Washington University in St. Louis
2013-2016
Meta (United States)
2016
Menlo School
2016
Aberystwyth University
2015
University of Electronic Science and Technology of China
2009-2011
Northeast Normal University
2010
Wheelock College
2004
With the advent of deep learning, neural network-based recommendation models have emerged as an important tool for tackling personalization and tasks. These networks differ significantly from other learning due to their need handle categorical features are not well studied or understood. In this paper, we develop a state-of-the-art model (DLRM) provide its implementation in both PyTorch Caffe2 frameworks. addition, design specialized parallelization scheme utilizing parallelism on embedding...
Abstract Hybrid glasses connect the emerging field of metal-organic frameworks (MOFs) with glass formation, amorphization and melting processes these chemically versatile systems. Though inorganic zeolites collapse around transition melt at higher temperatures, relationship between has so far not been investigated. Here we show how heating MOFs zeolitic topology first results in a low density ‘perfect’ glass, similar to those formed ice, silicon disaccharides. This order–order leads...
Convolutional neural networks (CNN) are increasingly used in many areas of computer vision. They particularly attractive because their ability to "absorb" great quantities labeled data through millions parameters. However, as model sizes increase, so do the storage and memory requirements classifiers, hindering applications such image speech recognition on mobile phones other devices. In this paper, we present a novel net- work architecture, Frequency-Sensitive Hashed Nets (FreshNets), which...
Training neural network language models over large vocabularies is computationally costly compared to count-based such as Kneser-Ney.We present a systematic comparison of strategies represent and train vocabularies, including softmax, hierarchical target sampling, noise contrastive estimation self normalization.We extend normalization be proper estimator likelihood introduce an efficient variant softmax.We evaluate each method on three popular benchmarks, examining performance rare words,...
Additive tree models (ATMs) are widely used for data mining and machine learning. Important examples of ATMs include random forest, adaboost (with decision trees as weak learners), gradient boosted trees, they often referred to the best off-the-shelf classifiers. Though capable attaining high accuracy, not well interpretable in sense that do provide actionable knowledge a given instance. This greatly limits potential on many applications such medical prediction business intelligence, where...
During the past decade, machine learning algorithms have become commonplace in large-scale real-world industrial applications. In these settings, computation time to train and test is a key consideration. At training-time must scale very large data set sizes.At testing-time, cost of feature extraction can dominate CPU runtime. Recently, promising method was proposed account for at testing time, called Cost-sensitive Tree Classifiers (CSTC). Although CSTC problem NP-hard, authors suggest an...
Convolutional neural networks (CNN) are increasingly used in many areas of computer vision. They particularly attractive because their ability to "absorb" great quantities labeled data through millions parameters. However, as model sizes increase, so do the storage and memory requirements classifiers. We present a novel network architecture, Frequency-Sensitive Hashed Nets (FreshNets), which exploits inherent redundancy both convolutional layers fully-connected deep learning model, leading...
This paper introduces a supervised metric learning algorithm, called kernel density (KDML), which is easy to use and provides nonlinear, probability-based distance measures. KDML constructs direct nonlinear mapping from the original input space into feature based on estimation. The in embodies established measures between probability functions, leads correct classification datasets for linear methods would fail. It addresses severe challenge kNN when features are heterogeneous domains and,...
This paper introduces a nonlinear logistic regression model for classification. The main idea is to map the data feature space based on kernel density estimation. A discriminative then learned optimize weights as well bandwidth of Nadaraya-Watson estimator. We propose hierarchical optimization algorithm learning coefficients and bandwidths in an integrated way. Compared other models such (KLR) SVM, our approach far more efficient since it solves problem with much smaller size. Two major...
A series of mesoporous CeZrTiOx catalysts were prepared by a facile hydrothermal method. Compared with CeTiOx synthesized under the same conditions, catalytic activity and anti-SO2 performance Ce1Zr1TiOx catalyst are greatly improved, at gas hourly space velocity (GHSV) 60 000 h-1, NOx removal efficiency is maintained 90% in temperature range 290-500 °C. The effect ZrO2 on Ce-Ti NH3-SCR was elucidated through characterizations. results revealed that doping Zr could significantly improve...
A series of Al2O3–CeO2 carriers were synthesized by hydrothermal method, and CuO/Al2O3–CeO2 catalysts prepared ultrasound-assisted impregnation for the catalytic oxidation CO C3H8. These samples have been characterized XRD, BET, TEM, XPS, other techniques. The 15 wt % CuO/A1C1 catalyst exhibited best activity, light-off temperatures (T50) C3H8 67 325 °C, respectively. XRD results showed that dispersion CuO on surface was improved introduction CeO2 into CuO/Al2O3 catalyst. Besides, with...
In traditional fractal image compression, the encoding procedure is time-consuming due to full search mechanism. order speedup encoder, we adopt particle swarm optimization method performed under classification and Dihedral transformation further decrease amount of MSE computations. The classifier partitions all blocks in domain pool range into three classes according third level wavelet coefficients. Each block searches most similar only from same class. Furthermore, property...
Metric learning, the task of learning a good distance metric, is key problem in machine with ample applications. This paper introduces novel framework for nonlinear metric called kernel density (KDML), which easy to use and provides nonlinear, probability-based measures. KDML constructs direct mapping from original input space into feature based on estimation. The embodies established measures between probability functions, leads accurate classification datasets existing linear methods would...
In this paper, we propose a novel supervised learning method, Fast Flux Discriminant (FFD), for large-scale nonlinear classification. Compared with other existing methods, FFD has unmatched advantages, as it attains the efficiency and interpretability of linear models well accuracy models. It is also sparse naturally handles mixed data types. works by decomposing kernel density estimation in entire feature space into selected low-dimensional subspaces. Since there are many possible...
Novel Ce x Zr1-x O2 (x = 0.67, 0.8, 0.9, 1.0) catalysts were designed and synthesized by solvothermal, calcination, sol-gel methods used to catalyze oxidation of soot from diesel vehicle exhaust. The influence different Ce/Zr molar ratios on the performance was investigated. These characterized XRD, N2 adsorption-desorption, FT-IR, TEM, XPS, H2-temperature programmed reduction (TPR), O2-temperature desorption (TPD) techniques. results indicated that Ce0.8Zr0.2O2 prepared calcination method...
In this article, we have synthesized two new heteropolytungstate-based compounds [EMIM]4[SiW12O40] (1) and [EMIM]6[P2W18O62] · 4H2O (2) using the ionic liquid (IL) [EMIM]Br (EMIM = 1-ethyl-3-methylimidazolium) as a solvent characterized them by infrared (IR) ultraviolet (UV) spectra, thermogravimetric (TG) elemental analyses, electrochemistry, single-crystal X-ray analyses. Compound 1 is constructed from one [SiW12O40]4− four [EMIM]+. structure, [EMIM]+ are connected hydrogen bonds with...
OF THE DISSERTATION Learning with Scalability and Compactness by Wenlin Chen Doctor of Philosophy in Computer Science Washington University St. Louis, 2016 Professor Yixin Chen, Chair Artificial Intelligence has been thriving for decades since its birth. Traditional AI features heuristic search planning, providing good strategy tasks that are inherently searchbased problems, such as games GPS searching. In the meantime, machine learning, arguably hottest subfield AI, embraces data-driven...