A5069193300- Topic Modeling
- Natural Language Processing Techniques
- Face recognition and analysis
- 3D Shape Modeling and Analysis
- Biomedical Text Mining and Ontologies
- Low-power high-performance VLSI design
- Computer Graphics and Visualization Techniques
- Hand Gesture Recognition Systems
- EEG and Brain-Computer Interfaces
- VLSI and Analog Circuit Testing
- Parallel Computing and Optimization Techniques
- Blind Source Separation Techniques
- Radiation Effects in Electronics
- Machine Learning in Healthcare
- Mycobacterium research and diagnosis
- Data Analysis with R
- Embedded Systems and FPGA Design
- Manufacturing Process and Optimization
- Geotechnical Engineering and Underground Structures
- Human Pose and Action Recognition
- Product Development and Customization
- Statistical Methods in Epidemiology
- Advancements in Semiconductor Devices and Circuit Design
- Meta-analysis and systematic reviews
- Human Motion and Animation
National Institute of Advanced Industrial Science and Technology
2023-2024
University of Electro-Communications
2021-2023
Nara Institute of Science and Technology
2017-2018
Tokyo Institute of Technology
2018
Initiatives (Denmark)
2018
Advanced Telecommunications Research Institute International
2015-2016
Fukuoka University
2007-2013
Hitachi (Netherlands)
2011
Hiroshima University
2007-2009
One difficulty in performing meta-analyses of observational cohort studies is that the availability confounders may vary between cohorts, so some cohorts provide fully adjusted analyses while others only partially analyses. Commonly, association an exposure and disease either are restricted to with full confounder information, or use all but do not adjust for confounding. We propose using a bivariate random-effects meta-analysis model information from available still adjusting potential...
Many measures have been proposed to summarize the prognostic ability of Cox proportional hazards (CPH) survival model, although none is universally accepted for general use. By contrast, little work has done stratified CPH model; such would be useful in analyses individual participant data from multiple studies, multi-centre and single study analysis where stratification used avoid making assumptions hazards. We chosen three developed unstratified model (Schemper Henderson's V , Harrell's...
Because of recent replacement physical documents with electronic medical records (EMR), the importance information processing in field has increased. In light this trend, we have been developing MedEx/J, which retrieves important Japanese language from reports. MedEx/J executes two tasks simultaneously: (1) term extraction, and (2) positive negative event classification. We designate approach as a one-scan approach, providing simplicity systems reasonable accuracy. performance on is...
高度な人工知能研究のためには,その材料となるデータが必須となる.医療,特に臨床に関わる分野において,人工知能研究の材料となるデータは主に自然言語文を含む電子カルテである.このようなデータを最大限に利用するには,自然言語処理による情報抽出が必須であり,同時に,情報抽出技術を開発するためのコーパスが必要となる.本コーパスの特徴は,45,000 テキストという我々の知る限りもっとも大規模なデータを構築した点と,単に用語のアノテーションや用語の標準化を行っただけでなく,当該の疾患が実際に患者に生じたかどうかという事実性をアノテーションした点の 2...
We introduce a novel parameterization of facial expressions by using elastic surface model. The model has been used as deformation tool especially for nonrigid organic objects. parameter is either retrieved from existing articulated face models or obtained indirectly manipulating muscles. can be applied on target dissimilar to the source create expressions. Due limited number control points, animation data created require less storage size without affecting range it provides. proposed method...
This paper proposes a novel fixed low-rank spatial filter estimation for brain computer interface (BCI) systems with an application that recognizes emotions elicited by movies. The proposed approach unifies such tasks as feature extraction, selection, and classification, which are often independently tackled in "bottom-up" manner, under regularized loss minimization problem. function is explicitly derived from the conventional BCI solves its optimization nonconvex constraint. For evaluation,...
In this paper, we propose a fixed low-rank spatial filter estimation for brain computer interface (BCI) systems with an application that recognizes emotions elicited by movies. The proposed approach unifies such tasks as feature extraction, selection, and classification, which are often independently tackled in "bottom-up" manner, under regularized loss minimization problem. We explicitly derive the function from conventional BCI solve its optimization non-convex constraint. For evaluation,...
The demand of power saving and highly dependable LSI has increased by the miniaturization device process technology spread portable devices such as mobile phones. design method which takes worst case scenario makes margin too large because parameter variations in deep submicron domain it serious impact for performance consumption. To deal with excessive margins, typical-case canary FF been proposed so far. By using FF, variability-aware guard band can be decreased. In this paper, we describe...
Soft error rate (SER) of various radiation hardened latches is analyzed by simulation. SER estimated modeling the variety current pulses triggered particle strikes such as neutrons from space or alpha particles using Monte Carlo method. By proposed method, we show that accurately without conducting irradiation experiments. As for soft tolerant latches, use DICE, Path-exclusive, BISER and TMR latches. The DICE Path-exclusive achieve resilience adopting local redundancy with feedbacks cross...
MultiCore processor is one of the promising techniques to satisfy computing demands future consumer devices. However, still threatened by increasing energy consumption due PVT (Process-Voltage-Temperature) variations. They require large design margins in supply voltage, resulting consumption. The combination DVS (Dynamic voltage scaling) technique and Canary FF (flip-flop), named Canary-DVS, has been proposed eliminate overestimated margin but only evaluated under assumption typical delay....
We propose an 'end-to-end' character-based recurrent neural network that extracts disease named entities from a Japanese medical text and simultaneously judges its modality as either positive or negative; i.e., the mentioned symptom is affirmed negated. The motivation to adopt networks learn effective lexical structural representation features for Entity Recognition also Positive/Negative classification annotated corpora without explicitly providing any rule-based manual feature sets....
<title>Abstract</title> <bold>Background</bold>: Named entity recognition (NER) aims to detect mentions from text and classify them into predefined types. It is a fundamental task in information extraction many other downstream tasks. However, the necessity of extensive human efforts annotate large amount training data imposes restrictions on state-of-the-art supervised deep learning methods, especially specific domains. To address this problem, various distantly methods (DS-NER), which aim...
We propose a novel pipeline method for translating signed Japanese sentences into written Japanese. Sign languages often suppress functional words such as particles, and most are not morphologically inflected they in spoken languages. Our explicitly compares contrasts the two divides translation process tasks: first, it translates glosses lemmatized or phrases, followed by complementing particles conjugating predicates verbs, auxiliary adjectives. is especially effective when size of...
We propose a distantly supervised pipeline NER which executes entity span detection and classification in sequence named DISTANT (DIstantly Supervised enTity spAN deTection classification).The former detector extracts possible mention spans by the distant supervision. Then later classifier assigns each to one of positive types or none employing unlabeled (PU) learning framework. Two models were built based on pre-trained SciBERT model fine-tuned with silver corpus generated...
We describe a novel NURBS surface fitting method to human face model. Our first step is resample of the original mesh The next we fit tensor product grid points. characteristic in that use only one patch By eliminating need make boundaries, our resampling and can be used for interactive system with minimum user interactions.
In this paper the authors investigated experimentally on stress distribution and flexual rigidity of a simply supported long pipe beam containing water under several different levels.It was ascertained that S. Timoshenko's theory is not suitable for adaption to thin shell structure with large deformation.The introduced useful approximate formula practical design kind structure.