- Text and Document Classification Technologies
- Domain Adaptation and Few-Shot Learning
- Advanced Graph Neural Networks
- Digital and Cyber Forensics
- Crime Patterns and Interventions
- Data Visualization and Analytics
- Topic Modeling
- Image Retrieval and Classification Techniques
- Advanced Image and Video Retrieval Techniques
- Distributed systems and fault tolerance
- Natural Language Processing Techniques
- Data Management and Algorithms
- Surface Roughness and Optical Measurements
- Machine Learning and ELM
- Data Quality and Management
- Machine Learning and Data Classification
- Korean Peninsula Historical and Political Studies
- Petri Nets in System Modeling
- Gaussian Processes and Bayesian Inference
- Gene expression and cancer classification
- Face and Expression Recognition
- Recommender Systems and Techniques
- Video Analysis and Summarization
- Real-Time Systems Scheduling
- Industrial Vision Systems and Defect Detection
Institut national de recherche en informatique et en automatique
2023-2024
Inria Saclay - Île de France
2023-2024
Sorbonne Université
2023
Pai Chai University
2002
Korea Advanced Institute of Science and Technology
2000-2001
Research Institute of Industrial Science and Technology
1991
C.L. Liu and J.W. Layland (1973) provided a utilization bound test which is applicable to rate-monotonic priority assignment, when all tasks deadlines are at the end of their periods. Subsequently, J. Lehoczky et al. (1989) an exact schedulability criterion any arbitrary assignment scheme, with no restriction on task deadlines, can be used computation times known exactly. In this work, we fill important gap between these two, by presenting technique for deriving bounds, based linear...
Most of the content-based image retrieval systems require a distance computation for each candidate in database. As brute-force approach, exhaustive search can be employed this computation. However, is time-consuming and limits usefulness such systems. Thus, there growing demand fast algorithm which provides same results as search. We propose based on multiresolution data structure. The proposed computes lower bound at level compares it with latest minimum distance, starting from...
Machine learning approaches have been introduced to support criminal investigations in recent years. In investigations, Criminal acts may be similar, and similar incidents occur consecutively by the same offender or group. Among various machine algorithms, network-based algorithms will suitable reflect such associations. general, however, inference is slow when size of data large, so it fatal crime scenes that require urgency. And worse, network must able handle complex information entangled...
Graph-based models have gained much interest in the domain of machine learning as they offer advantage handling data that reside on complex structures. From various encounter graph-structured data, graph-based semi-supervised (SSL) shown successful results multiple applications. The key idea behind SSL is spreading process labels through edges and problem boils down to keeping graph Laplacian intact. Meanwhile, with rapid growth availability there exist descriptions graphs for same set...
Autoencoders are widely used for dimensionality reduction nonlinearly. However, determining the number of nodes in autoencoder embedding space is still a challenging task. The bottleneck layer, which an encoded representation, estimated and determined by users. Therefore, to maintain performance reduce complexity model, indicator that automatically selects needed. This study proposes method estimating adequate layer while training model. basic idea proposed eliminate lazy rarely affect model...
One of the interesting characteristics crime data is that criminal cases are often interrelated. Criminal acts may be similar, and similar incidents occur consecutively by same offender or group. Among many machine learning algorithms, network-based approaches well-suited to reflect these associative characteristics. Applying networks composed their associates can predict potential suspects. This narrows scope an investigation, saving time cost. However, inference from not straightforward as...
There are increasingly efficient data processing pipelines that work on vectors of numbers, for instance most machine learning models, or vector databases fast similarity search. These require converting the to numbers. While this conversion is easy simple numerical and categorical entries, strife with text such as names descriptions. In age large language what's best strategies vectorize tables baring in mind larger models entail more operational complexity? We study benefits 14 analytical...
This paper presents a fast search algorithm based on multi- resolution data structure for efficient image retrieval in large databases. The proposed consists of two stages: database-building stage and searching stage. In the stage, we partition set into pre-defined number clusters by using MacQueen K-means clustering algorithm. has steps to choose proper find best match among all images included chosen clusters. order reduce heavy computational cost kinds exhaustive algorithms feature space,...
The classification of complicated 2-D shapes such as cracks fractures and the abnormal patterns found on surfaces materials is very important in automated surface inspection. As a method facilitating these we report strategy extracting hierarchical structures from complex using branch cutting boundary analysis others. Experimental results are given with random statistical images taken some materials.