A5100660677- Software Engineering Research
- AI in cancer detection
- Writing and Handwriting Education
- Digital Imaging for Blood Diseases
- Educational Technology and Assessment
- Medical Image Segmentation Techniques
- Advanced Neural Network Applications
- Image and Signal Denoising Methods
- Topic Modeling
- Natural Language Processing Techniques
- Advanced Image Processing Techniques
- Intelligent Tutoring Systems and Adaptive Learning
- Advanced Image Fusion Techniques
- Text Readability and Simplification
- Psychometric Methodologies and Testing
- Speech Recognition and Synthesis
- Cryptographic Implementations and Security
- Radiomics and Machine Learning in Medical Imaging
- Domain Adaptation and Few-Shot Learning
- Higher Education and Teaching Methods
- COVID-19 diagnosis using AI
- Distributed and Parallel Computing Systems
- Digital Communication and Language
- Multi-Criteria Decision Making
- User Authentication and Security Systems
Educational Testing Service
2013-2024
Peking University
2018-2022
Beijing Institute of Big Data Research
2019-2022
University of Birmingham
2022
Research Institute of Highway
2021
Ministry of Transport
2021
Chongqing University of Posts and Telecommunications
2021
Beijing Technology and Business University
2021
Centre National de la Recherche Scientifique
2016-2018
Institut d'Électronique et des Technologies du numéRique
2016-2018
Unsupervised domain adaption has recently been used to reduce the shift, which would ultimately improve performance of semantic segmentation on unlabeled real-world data. In this paper, we follow trend propose a novel method shift using strategies discriminator attention and self-training. The strategy contains two-stage adversarial learning process, explicitly distinguishes well-aligned (domain-invariant) poorly-aligned (domain-specific) features, then guides model focus latter....
Red blood cell (RBC) segmentation and classification from microscopic images is a crucial step for the diagnosis of sickle disease (SCD). In this work, we adopt deep learning based semantic framework to solve RBC task. A major challenge robust large variations on size, shape viewpoint cells, combining with low image quality caused by noise artifacts. To address these challenges, apply deformable convolution layers classic U-Net structure implement (dU-Net). architecture has been shown offer...
Simultaneous modeling of the spatio-temporal variation patterns brain functional network from 4D fMRI data has been an important yet challenging problem for field cognitive neuroscience and medical image analysis. Inspired by recent success in applying deep learning decoding encoding, this work we propose a convolutional neural (ST-CNN)to jointly learn spatial temporal targeted training perform automatic, pin-pointing identification. The proposed ST-CNN is evaluated task identifying Default...
Sparse deconvolution methods frequently invert for subsurface reflection impulses and adopt a trace-by-trace processing pattern. However, following this approach causes unreliability of the estimated reflectivity due to nonuniqueness inverse problem, poor spatial continuity structures in reconstructed section, suppression on signals with small amplitudes. We have developed structurally constrained multichannel band-controlled (SC-MBCD) algorithm alleviate these three issues. The inverts...
Analysis of keystroke logging data is increasing interest, as evident from a substantial amount recent research on the topic. Some has focused prediction essay scores features, but linear regression only method that been used in this research. Data mining methods such boosting and random forests have found to improve over traditional various scientific fields, not features. This article first provides review boosting, which popular method. The then applies predict large number features other...
This paper presents a multidimensional model of variation in writing quality, register, and genre student essays, trained tested via confirmatory factor analysis 1.37 million essay submissions to ETS' digital service, Criterion®. The was also validated with several other corpora, which indicated that it provides reasonable fit for data from 4th grade college. It includes an the test‐retest reliability each trait, longitudinal trends by both within school year 12th grades, differences using...
Reliable cell segmentation and classification from biomedical images is a crucial step for both scientific research clinical practice. A major challenge more robust methods the large variations in size, shape viewpoint of cells, combining with low image quality caused by noise artifacts. To address this issue, work we propose learning-based, simultaneous method based on deep U-Net structure deformable convolution layers. The architecture learning has been shown to offer precise localization...
Cell detection and cell type classification from biomedical images play an important role for high-throughput imaging various clinical application. While of single sample can be performed with standard computer vision machine learning methods, analysis multi-label samples (region containing congregating cells) is more challenging, as separation individual cells difficult (e.g. touching or even impossible overlapping cells). As multi-instance are common in analyzing Red Blood (RBC) Sickle...
ABSTRACT Many testing programs use automated scoring to grade essays. One issue in essay that has not been examined adequately is population invariance and its causes. The primary purpose of this study was investigate the impact sampling model calibration on scores. This analyzed scores produced by e‐rater ® engine using a GRE assessment data set. Results suggested equal allocation stratification language approach performed optimally maximizing either human/e‐rater agreement or their...
This study investigates the effects of a scenario-based assessment design on students' writing processes. An experimental data set consisting four conditions was used in which number scenarios (one or two) and placement essay task with respect to lead-in tasks (first vs. last) were varied. Students' processes recorded using keystroke logs. Each action classified into one states: planning, text production, local edit, jump semi-Markov model fit data. Results showed that single-scenario...
Existing object detection algorithms are affected by scenes with poor robustness, besides the existing public datasets not applicable to urban road traffic scenes. In order solve problems of low accuracy in detecting panoramic video images, high false rate, this article designed a real-time information method based on multi-scale feature fusion. For start, vehicle equipped hp-f515 driving recorder collected under real scene Beijing. The total length route was 11 km. Extracted recorded video,...
Breast cancer is one of the most common cancers worldwide, and early detection can significantly reduce its mortality rate. It crucial to take multi-scale information tissue structure into account in breast cancer. And thus, it key design an accurate computer-aided (CAD) system capture contextual features a cancerous tissue. In this work, we present novel graph convolutional neural network for histopathological image classification The new method, named wavelet (MS-GWNN), leverages...
Abstract Automated essay scoring ( AES ) generally computes scores as a function of macrofeatures derived from set microfeatures extracted the text using natural language processing NLP ). In e‐rater ® automated engine, developed at Educational Testing Service ETS for essays, each four grammar, usage, mechanics , and style [ GUMS ]) is computed microfeatures. Statistical analyses reveal that some these might not explain much variance in human regardless writing tasks. Currently, same...
Extracting multi-scale information is key to semantic segmentation. However, the classic convolutional neural networks (CNNs) encounter difficulties in achieving extraction: expanding kernel incurs high computational cost and using maximum pooling sacrifices image information. The recently developed dilated convolution solves these problems, but with limitation that dilation rates are fixed therefore receptive field cannot fit for all objects different sizes image. We propose an...
As a new extended model of fuzzy sets, hesitant set theory is useful tool to process uncertain information in decision making problems. The traditional multi-attribute (MADM) can only choose an optimal strategy, which not suitable for all the complex scenarios. Typically, practical application, problems may be more complicated involving three options acceptance, non-commitment and rejection decisions. Three-way decisions, divide universe into disjoint regions by pair thresholds, are...