A5075541712- Data Management and Algorithms
- Advanced Database Systems and Queries
- Recommender Systems and Techniques
- Advanced Image and Video Retrieval Techniques
- Advanced Graph Neural Networks
- Adversarial Robustness in Machine Learning
- Advanced Data Storage Technologies
- Topic Modeling
- Advanced Bandit Algorithms Research
- Catalysis for Biomass Conversion
- Lignin and Wood Chemistry
- Algorithms and Data Compression
- Privacy-Preserving Technologies in Data
- Spam and Phishing Detection
- Machine Learning and Algorithms
- Image Retrieval and Classification Techniques
- Catalysis and Hydrodesulfurization Studies
- Stochastic Gradient Optimization Techniques
- Natural Language Processing Techniques
- Gamma-ray bursts and supernovae
- Scientific Computing and Data Management
- Data Mining Algorithms and Applications
- Advanced Neural Network Applications
- Planetary Science and Exploration
- Multimodal Machine Learning Applications
Rochester Institute of Technology
2022-2024
Shenzhen Institutes of Advanced Technology
2024
Chinese Academy of Sciences
2019-2024
University of Chinese Academy of Sciences
2022-2023
Institute of Engineering Thermophysics
2022-2023
LinkedIn (United States)
2023
Bellevue Hospital Center
2021-2022
Baidu (China)
2020-2022
Yangtze University
2022
Cognitive Research (United States)
2020-2021
Merging neutron stars offer an exquisite laboratory for simultaneously studying strong-field gravity and matter in extreme environments. We establish the physical association of electromagnetic counterpart EM170817 to gravitational waves (GW170817) detected from merging stars. By synthesizing a panchromatic dataset, we demonstrate that are long-sought production site forging heavy elements by r-process nucleosynthesis. The weak gamma-rays seen dissimilar classical short gamma-ray bursts with...
Recently, machine learning models have demonstrated to be vulnerable backdoor attacks, primarily due the lack of transparency in black-box such as deep neural networks. A third-party model can poisoned that it works adequately normal conditions but behaves maliciously on samples with specific trigger patterns. However, injection function is manually defined most existing attack methods, e.g., placing a small patch pixels an image or slightly deforming before poisoning model. This results...
A novel process for efficient <italic>para</italic> ethyl phenol production was provided <italic>via</italic> the selective cleavage of <italic>p</italic>CA8 linkage lignin.
As one of the major search engines in world, Baidu's Sponsored Search has long adopted use deep neural network (DNN) models for Ads click-through rate (CTR) predictions, as early 2013. The input futures used by online advertising system (a.k.a. "Phoenix Nest'') are extremely high-dimensional (e.g., hundreds or even thousands billions features) and also sparse. size CTR production can well exceed 10TB. This imposes tremendous challenges training, updating, using such production. For system,...
Neural networks of ads systems usually take input from multiple resources, e.g., query-ad relevance, ad features and user portraits. These inputs are encoded into one-hot or multi-hot binary features, with typically only a tiny fraction nonzero feature values per example. Deep learning models in online advertising industries can have terabyte-scale parameters that do not fit the GPU memory nor CPU main on computing node. For example, sponsored system contain more than $10^{11}$ sparse making...
Bulk chemicals produced from renewable resources are receiving considerable attention due to the increase in sustainable practices. In this study, a novel, efficient, and strategy has been proposed produce versatile petroleum-based monophenols natural aromatic polymer, lignin, via selective cleavage of its specific chemical linkages by cost-effective catalyst, Ni/MgO. The results demonstrate that 93.4% lignin is converted, yielding 15.0% presence 20% Importantly, 42.3% these volatile found...
Approximate nearest neighbor (ANN) searching is a fundamental problem in computer science with numerous applications (e.g.,) machine learning and data mining. Recent studies show that graph-based ANN methods often outperform other types of algorithms. For typical methods, the algorithm executed iteratively execution dependency prohibits GPU adaptations. In this paper, we present novel framework decouples on graph into 3 stages, order to parallel performance-crucial distance computation....
In the summer of 2003, a heat wave swept Europe and caused more than 70 000 additional fatalities [J.-M. Robine et al., C. R. Biologies331 (2008)]. Global warming climate change is no longer prophecy to be fulfilled, as strong waves typhoons, well severe rainfalls, are becoming severe. Extreme weather events in world, especially over North America, widely studied frequently reported media. These may related ongoing change. this NSR forum, active researchers specialized field gather discuss...
Deep neural network has been adopted as the standard model to predict ads click-through rate (CTR) for commercial online advertising systems. Deploying an industrial scale system requires overcome numerous challenges, e.g., hundreds or thousands of billions input features and also training samples, which under cost budget can cause fundamental issues on storage, communication, speed. In this work, we present Baidu's industrial-scale practices how apply machine learning techniques address...
Gradient Boosting Decision Tree (GBDT) is one of the most popular machine learning models in various applications. However, traditional settings, all data should be simultaneously accessed training procedure: it does not allow to add or delete any instances after training. In this paper, we propose an efficient online framework for GBDT supporting both incremental and decremental learning. To best our knowledge, first work that considers in-place unified on GBDT. reduce cost, present a...
Large Language Models (LLMs) are vulnerable to attacks like prompt injection, backdoor attacks, and adversarial which manipulate prompts or models generate harmful outputs. In this paper, departing from traditional deep learning attack paradigms, we explore their intrinsic relationship collectively term them Prompt Trigger Attacks (PTA). This raises a key question: Can determine if is benign poisoned? To address this, propose UniGuardian, the first unified defense mechanism designed detect...
Abstract This study focuses on the Dongzhuang and Sanmenxia reservoirs, addressing key issues such as water-sediment relationship in reservoir groups, reducing Tongguan gauge elevation, ensuring flood safety lower Weihe River. Utilizing hydrological data from significant stations Yellow River’s tributaries main stream spanning 1960 to 2020, employed Mann-Kendall statistical methods analyze evolution of characteristics relationships basin. The results indicate a declining trend both runoff...
It has been more than 10 years since Satoshi Nakamoto published his famous paper entitled 'Bitcoin: a peer-to-peer electronic cash system', which set the foundation of blockchain technology. Accompanied by price volatility bitcoins from 2017 to 2018, hot word on internet, and particularly in China. Blockchain offers distributed secure system for data storage value transactions. Its applications are springing up multiple fields. The Chinese government is considering these trends with great...
The 2019 Nobel Prize in Chemistry was awarded to three pioneers of lithium-ion batteries (LIBs)-Prof. John B. Goodenough at the University Texas, Prof. M. Stanley Whittingham State New York and Mr. Akira Yoshino Asahi Corporation Japan, which is a great encouragement whole field. LIBs have been developed for several decades with progress slowing down their performances approaching some theoretical limits. On other hand, new types or power systems, including solid-state batteries, sodium-ion...
Open AccessCCS ChemistryCOMMUNICATION1 Mar 2022Dynamic Macro- and Microgels Driven by Adenosine Triphosphate-Fueled Competitive Host–Guest Interaction Xiang Hao, Hairong Wang, Weijie Zhao, Liteng Feng Peng Qiang Yan Hao *Corresponding authors: E-mail Address: [email protected] Beijing Key Laboratory of Lignocellulosic Chemistry, Advanced Innovation Center for Tree Breeding Molecular Design, Forestry University, 100083 , Wang Zhao State Engineering Polymers, Fudan Shanghai 200433...
Along with the evolution of deep neural networks (DNNs) in many real-world applications, complexity model building has also dramatically increased. Therefore, it is vital to protect intellectual property (IP) builder and ensure trustworthiness deployed models. Meanwhile, adversarial attacks on DNNs (e.g., backdoor poisoning attacks) that seek inject malicious behaviors have been investigated recently, demanding a means for verifying integrity users. This paper presents novel DNN...
User-Defined Functions (UDF) allow application programmers to specify analysis operations on data, while leaving the data management tasks system. This general approach enables numerous custom functions and is at heart of modern Big Data systems. Even though UDF mechanism can theoretically support arbitrary operations, a wide variety common -- such as computing moving average time series, vorticity fluid flow, etc., are hard express slow execute. Since these traditionally performed...
Abstract There has been speculation about a class of relativistic explosions with an initial Lorentz factor Γ init smaller than that classical gamma-ray bursts (GRBs). These “dirty fireballs” would lack prompt GRB emission but could be pursued via their optical afterglow, appearing as transients fade overnight. Here we report search for such (that by 5- σ in magnitude overnight) four years archival photometric data from the intermediate Palomar Transient Factory (iPTF). Our criteria yielded...