A5100663869- Domain Adaptation and Few-Shot Learning
- Advanced Neural Network Applications
- Multimodal Machine Learning Applications
- Topic Modeling
- COVID-19 diagnosis using AI
- Reinforcement Learning in Robotics
- Advanced Graph Neural Networks
- IoT and Edge/Fog Computing
- Air Quality Monitoring and Forecasting
- Recommender Systems and Techniques
- Language, Metaphor, and Cognition
- Infection Control and Ventilation
- Machine Learning and ELM
- Infrastructure Maintenance and Monitoring
- Anomaly Detection Techniques and Applications
- Forest Management and Policy
- Building Energy and Comfort Optimization
- Educational Robotics and Engineering
- Ferroelectric and Negative Capacitance Devices
- Philosophy and Historical Thought
- Nonlinear Dynamics and Pattern Formation
- Machine Learning in Healthcare
- Sustainable Agricultural Systems Analysis
- Stochastic Gradient Optimization Techniques
- Fault Detection and Control Systems
City University of Hong Kong
2019-2023
Southwest University
2023
Texas A&M University
2023
Harbin Institute of Technology
2022
Beijing University of Civil Engineering and Architecture
2022
Huaneng Clean Energy Research Institute
2022
Shandong University
2022
Wuhan University of Technology
2020-2021
Alibaba Group (United States)
2021
Tianjin University
2020
This research aims to evaluate various traditional or deep machine learning algorithms for the prediction of groundwater level (GWL) using three key input variables specific Izeh City in Khuzestan province Iran: extraction rate (E), rainfall (R), and river flow (P) (with 3 km distance). Various (DML) algorithms, including convolutional neural network (CNN), recurrent (RNN), support vector (SVM), decision tree (DT), random forest (RF), generative adversarial (GAN), were evaluated. The (CNN)...
The recent breakthroughs in computer vision have benefited from the availability of large representative datasets (e.g. ImageNet and COCO) for training. Yet, robotic poses unique challenges applying visual algorithms developed these standard due to their implicit assumption over non-varying distributions a fixed set tasks. Fully retraining models each time new task becomes available is infeasible computational, storage sometimes privacy issues, while naïve incremental strategies been shown...
Current Reinforcement Learning (RL) methods often suffer from sample-inefficiency, resulting blind exploration strategies that neglect causal relationships among states, actions, and rewards. Although recent approaches aim to address this problem, they lack grounded modeling of reward-guided understanding states actions for goal-orientation, thus impairing learning efficiency. To tackle issue, we propose a novel method named Causal Information Prioritization (CIP) improves sample efficiency...
With the rapid advancement of Large Language Models (LLMs), safety LLMs has been a critical concern requiring precise assessment. Current benchmarks primarily concentrate on single-turn dialogues or single jailbreak attack method to assess safety. Additionally, these have not taken into account LLM's capability identifying and handling unsafe information in detail. To address issues, we propose fine-grained benchmark SafeDialBench for evaluating across various attacks multi-turn dialogues....
Recent breakthroughs in computer vision areas, ranging from detection, segmentation, to classification, rely on the availability of large-scale representative training datasets. Yet, robotic poses new challenges towards applying visual algorithms developed these datasets because latter implicitly assume a fixed set categories and time-invariant distribution tasks. In practice, assistive robots should be able operate dynamic environments with everyday changes. The variations four commonly...
Conversational Recommender Systems (CRSs) in E-commerce platforms aim to recommend items users via multiple conversational interactions. Click-through rate (CTR) prediction models are commonly used for ranking candidate items. However, most CRSs suffer from the problem of data scarcity and sparseness. To address this issue, we propose a novel knowledge-enhanced deep cross network (K-DCN), two-step (pretrain fine-tune) CTR model We first construct billion-scale conversation knowledge graph...
Humans have a remarkable ability to learn continuously from th e environment and inner experience. One of the grand goals robots is build an artificial "lifelong learning" agent that can shape cultivated understanding world current scene previous knowledge via autonomous lifelong development. It challenging for robot learning process retain earlier when encounter new tasks or information. Recent advances in computer vision deep -learning methods been impressive due large-scale data sets,...
The contactless estimation of the weight a container and amount its content manipulated by person are key pre-requisites for safe human-to-robot handovers. However, opaqueness transparencies content, variability materials, shapes, sizes, make this difficult. In paper, we present range methods an open framework to benchmark acoustic visual perception capacity container, type, mass, content. includes dataset, specific tasks performance measures. We conduct in-depth comparative analysis that...
Fingerprint images are textural consisting of ridges and valleys. The orientation textures can be determined by field computation. is the critical basis for fingerprint image segmentation, filtering enhancement matching processes, algorithm plays a very important role in applied Automated Identification Systems (AFIS). available algorithms include mainly mask gradient algorithm. They used spatial frequency domains, respectively, both offer satisfactory matrixes. Based on an experimental...
Pattern formations in an Oregonator model with superdiffusion are studied two-dimensional (2D) numerical simulations. Stability analyses performed by applying Fourier and Laplace transforms to the space fractional reaction–diffusion systems. Antispiral, stable turing patterns, travelling patterns observed changing diffusion index of activator. Analyses Floquet multipliers show that limit cycle solution loses stability at wave number primitive vector hexagonal pattern. We also a transition...
Personnel search and rescue work is a kind of emergency mainly to save the lives personnel. If only firefighters carry out in unknown areas field, efficiency too low it easy delay best time. joint carried using mobile robots firefighters, will be beneficial find missing persons with vital signs timely manner, convenient soothe emotions family members embodying humanistic care humanitarian spirit. In order solve problem that current robot has high energy consumption short flight time, can not...
Abstract This work is devoted to bridging the gap between large‐area, economically driven macromodels such as C anadian R egional A griculture M odel ( CRAM ) and small‐area biophysically based process models used in environmental assessments through development of a L U se llocation LUAM ). designed enable economic scenarios be conducted by allocating crop area changes predicted for large areas much smaller S oil andscapes anada SLC polygons an optimization method on land capability,...
The recent breakthroughs in computer vision have benefited from the availability of large representative datasets (e.g. ImageNet and COCO) for training. Yet, robotic poses unique challenges applying visual algorithms developed these standard due to their implicit assumption over non-varying distributions a fixed set tasks. Fully retraining models each time new task becomes available is infeasible computational, storage sometimes privacy issues, while na\"{i}ve incremental strategies been...
With the development of Internet+medicine, online medical treatment has gradually become new direction industry. Many hospitals provide registration services to public, and due lack professional knowledge patients, problem wrong often occurs. How use deep learning technology help patients reduce waste resources an urgent problem. To address above problems, this paper proposes ERNIE-based text classification model for intelligent triage. The consists two parts, ERNIE BiGRU. pre-training is...
Training deep neural networks (DNNs) typically requires massive computational power. Existing DNNs exhibit low time and storage efficiency due to the high degree of redundancy. In contrast most existing DNNs, biological social with vast numbers connections are highly efficient scale-free properties indicative power law distribution, which can be originated by preferential attachment in growing networks. this work, we ask whether topology best performing shows similar how use construct...
General intelligence requires quick adaption across tasks. While existing reinforcement learning (RL) methods have made progress in generalization, they typically assume only distribution changes between source and target domains. In this paper, we explore a wider range of scenarios where both the environment spaces may change. For example, Atari games, train agents to generalize tasks with different levels mode difficulty, there could be new state or action variables that never occurred...
With the return of passengers to work, airport has a huge flow people at entrance and exit. The existing detection disinfection system is generally not intelligent enough, development passenger health information registration minimal, which often leads problems such as missing temperature measurement, time-consuming, low efficiency, difficulty in establishing databases. In order solve make more suitable for “strict prevention” “fine control” stages epidemic prevention control, this paper...