A5114804114- Context-Aware Activity Recognition Systems
- Face and Expression Recognition
- Machine Learning and ELM
- Industrial Vision Systems and Defect Detection
- Face recognition and analysis
- Photovoltaic System Optimization Techniques
- Healthcare Technology and Patient Monitoring
- IoT and Edge/Fog Computing
- Technology Use by Older Adults
- Mobile Health and mHealth Applications
- Advanced Bandit Algorithms Research
- Data Management and Algorithms
- Sparse and Compressive Sensing Techniques
- Innovative Human-Technology Interaction
- Usability and User Interface Design
- Distributed and Parallel Computing Systems
- Advanced Neural Network Applications
- Interconnection Networks and Systems
- Parallel Computing and Optimization Techniques
- Design Education and Practice
University of Science and Technology Beijing
2023
Tianjin University
2019
Google (United States)
2017
Hebei North University
2015
University of California, Riverside
2005
How to help patients find their departments or wards is an issue that needs be solved by hospitals urgently.This paper presents iBeacon-based indoor positioning system for hospitals.It firstly analyzes the advantages of iBeacon compared with common technologies; then designs based on three-layer architecture Internet things have message-push-service through clients.Finally, shortest distance algorithm Floyd used recommend nearest department ward patients.Shown as result experiment, can...
<title>Abstract</title> Chain-based reasoning methods like chain of thought (CoT) play a rising role in solving tasks for large language models (LLMs). However, the causal hallucinations between step and corresponding state transitions are becoming significant obstacle to advancing LLMs' capabilities, especially long-range tasks. This paper proposes non-chain-based framework simultaneous consideration significance consistency, i.e., Causal Significance Consistency Enhancer (CSCE). We...
In apparel recognition, deep neural network models are often trained separately for different verticals (e.g. [7]). However, using specialized is not scalable and expensive to deploy. This paper addresses the problem of learning one unified embedding model multiple object all classes) without sacrificing accuracy. The tackled from two aspects: training data difficulty. On aspect, we figure out that a single with triplet loss, there an accuracy sweet spot in terms how many together. To ease...
Online kernel selection is fundamental to online learning. In contrast offline selection, intermixes and training at each round of learning, requires a sublinear regret bound low computational complexity. this paper, we first compare the difference between then survey existing approaches from perspectives formulation, algorithm, candidate kernels, complexities guarantees, finally point out some future research directions in selection. This article categorized under: Technologies > Machine...
Using portable tools to monitor and identify daily activities has increasingly become a focus of digital healthcare, especially for elderly care. One the difficulties in this area is excessive reliance on labeled activity data corresponding recognition modeling. Labeled expensive collect. To address challenge, we propose an effective robust semi-supervised active learning method, which combines mainstream method with expert collaboration. Our takes user's trajectory as only input. In...
Human Activity Recognition (HAR) has become a subject of concern and plays an important role in daily life. HAR uses sensor devices to collect user behavior data, obtain human activity information identify them. Markov Logic Networks (MLN) are widely used as effective combination knowledge data. MLN can solve the problems complexity uncertainty, good expression ability. However, structure learning is relatively weak requires lot computing storage resources. Essentially, derived from data...
At a time when photovoltaic power generation is becoming more and mature, the maintenance of modules has also become major problem. In order to avoid environmental impact hot spot damage panels, this leads decrease in efficiency, as well current detection means small spots are often be missed, combined with need for speed accuracy UAV inspection cleaning, traditional two-level method, ResNet residual network-based full convolutional network model, further use grayscale, filtering edge...
Online kernel ridge regression via existing sampling approaches, which aim at approximating the matrix as accurately possible, is independent of learning and has a cubic time complexity with respect to size for updating hypothesis. In this paper, we propose new online an incremental predictive approach, nearly optimal accumulated loss performs efficiently each round. We use estimated leverage score labeled matrix, depends on round, construct distribution, probability Nyströ m approximation....