[Information and future] Year Started: 2016

*This Program finished in FY2021
 Active Period: FY2016 - FY2021
 Affiliations and titles are at the end of each research activity.

Yuichi Aoki

Development of pairwise deep learning architectures for the modeling of gene-gene interactions

Researcher

Yuichi Aoki

Yuichi Aoki

Tohoku University
Tohoku Medical Megabank Organization
Assistant Professor

Outline

This research aims to develop a new deep learning architecture for the modeling of gene-gene interactions, such as gene co-expression and protein-protein interactions. We try to modify the conventional deep neural network model to allow it: (1) to accept two inputs (two feature vectors for any gene pair) and (2) to capture important feature similarities automatically. The developed model will contribute not only to medical/agricultural applications such as genomic medicine and crop breeding, but also to knowledge acquisition from big data in various non-biological fields.

Yuki Arase

Grounded Phrase Embedding for Multimodal Information Processing

Researcher

Yuki Arase

Yuki Arase

Osaka University
Graduate school of information science and technology
Associate Professor

Outline

Text is not computable as is, since it is composed of symbols. This project aims to project text into numerical vectors embedding their syntactic structures, as well as ground the vectors based on their semantic relations. For example, a linguistic structure in “ripe yellow banana” represents that something ripped is banana. Also, we can easily come up with its paraphrase, “mellow banana.” We replicate such syntactic and semantic information on the vector. With this technology, language can be computed together with image and sensor vectors, which contributes to create robots with human-like recognition abilities.

Satoshi Iizuka

Automatic Restoration of Historical Video and Image Contents via Deep Neural Networks

Researcher

Satoshi Iizuka

Satoshi Iizuka

Waseda University
Faculty of Science and Engineering
Junior Researcher

Outline

I will establish a fully automatic architecture for restoring historical video and image contents via deep learning. Old photography usually consists of black and white images with noise and blur. Up until now, removing the deterioration and restoring the color required complicated image editing by an expert through trial and error. In this work, I will propose a novel architecture based on multiple convolutional neural networks to automatically detect time-related deterioration and restore them with color. I believe this will make people more familiar with historical image contents, which will improve the understanding of human history.

Yoshio Ishiguro

Interaction method for autonomous driving vehicles in real oriented virtuality environment

Researcher

Yoshio Ishiguro

Yoshio Ishiguro

Nagoya University
Institute of Innovation for Future Society
Research Associate Professor

Outline

This research propose a novel interaction system for autonomous driving vehicles. We create a “real oriented virtuality: RoV” environment that imports actual situations from a real environment and reconstructs the real environment into a virtual environment in real time. We design the real oriented virtuality environment for autonomous driving vehicles to reduce and highlight information about the actual environment, such as pedestrians, traffic signals, and other obstacles. Passengers can recognize the actual context of the autonomous driving behavior using reconstructed virtual environment.

Nakamasa Inoue

Higher-Order Models for Semantic Concept Discovery on Multimedia Data

Researcher

Nakamasa Inoue

Nakamasa Inoue

Tokyo Institute of Technology
Department of Computer Science
Assistant Professor

Outline

The goal of this project is to create higher-order models for discovering semantic concepts from multimedia data. Here, a higher-order model is a model to generate detection models for unseen semantic concepts such as objects, actions, and scenes. We develop a learning framework to obtain a higher-order model from pre-trained deep neural networks. We evaluate the framework on a large-scale video dataset by finding and detecting unseen objects, event and scenes.

Akira Imakura

Development of a nonlinear nonnegative matrix factorization-based algorithm for deep neural networks

Researcher

Akira Imakura

Akira Imakura

University of Tsukuba
Faculty of Engineering, Information and Systems
Assistant Professor

Outline

The backpropagation, based on the stochastic gradient descent method, is the most successful and the de-facto standard algorithm for computing deep neural networks (DNNs) in several applications such as image recognition. In this research project, we propose and develop a novel algorithm for DNNs based on a nonlinear nonnegative matrix factorization, which is compatible with parallel computing.

Akira Uchiyama

Development of Sustainable IoT Platform by Cooperative Battery-less Sensors

Researcher

Akira Uchiyama

Akira Uchiyama

Osaka University
Graduate School of Information Science and Technology
Assistant Professor

Outline

The number of IoT devices is expected to be enormous, causing huge effort to maintain batteries. In this research, I realize cooperation among battery-less sensors by combining energy harvesting and ambient backscatter which connects sensors wirelessly with tiny energy consumption. The goal is to develop technologies for “Sustainable IoT platform”, which combines energy harvested at multiple devices to satisfy application requirements.

Yuki Uranishi

Estimation of Geometric and Photometric Parameters of Structural Color Object

Researcher

Yuki Uranishi

Yuki Uranishi

Osaka University
Cybermedia Center
Associate Professor

Outline

Through this research project, I propose a method for estimating geometric and optical parameters of an object which has a structural coloration on its surface. The proposed method aims at estimating the parameters in real-environment and real-time. A light field camera is employed to capture a light field on the surface of target object. The shape and the coloration model of the object is estimated by the finding the local changes of the structural color pattern from the light field.

Masahiro Oda

Development of colonoscope navigation system using colon deformation model

Researcher

Masahiro Oda

Masahiro Oda

Nagoya University
Graduate School of Information Science
Assistant Professor

Outline

Many computer aided diagnosis systems for colon have been developed. However, clinically useful assistance system of colon treatments, which can be performed after diagnosis, has not been proposed. Treatments of colon are performed based on experience of individual physician. In this research, I develop an accurate navigation system for colonoscopic treatment. I estimate colonoscopt tip position using a colon deformation model, which can be achieved by using machine learning techniques that learn colon deformations caused by colonoscope insertions. I develop a software that shows treatment assisting information to physicians such as current colonoscope tip position and distance to polyps.

Yuki Kubo

Input Methods Based on Metapfer by using Ultra-small Devices

Researcher

Yuki Kubo

Yuki Kubo

University of Tsukuba
Graduate School of Systems and Information Engineering
Master Course Student

Outline

The goal of this project is to construct a system that allows the user to interact with IoTs naturally, anywhere, and anytime by using ultra-small devices such as smartwatches. To do this, this project develops input methods that utilizes the users’ mental model of the target IoTs. As the results, they can be remembered and performed easily by the users. In order to realize such system, this project investigates input methods based on metaphors. In addition, recognition technologies are developed by utilizing sensors of ultra-small devices.

Nahoko Kuroki

An interdisciplinary study of physical chemistry and information science for realizing CO2 free society

Researcher

Nahoko Kuroki

Nahoko Kuroki

Ochanomizu University
Graduate School of Humanities and Sciences
Ph.D. Course Student

Outline

Efficient CO2 capture is indispensable for solving the global warming problem while keeping/improving our living quality. After selective CO2 absorption ability was found in some ionic liquids, chemists have tried to optimize the ability by changing the combination of ions that govern the physicochemical nature of ionic liquids. In this study, using an interdisciplinary approach of physical chemistry and information science, I will develop an efficient prediction system that can design ionic liquids with higher CO2 absorption ability. With a perspective in the field of information science, I would like to realize CO2 free society.

Ryota Kobayashi

Towards automatic analysis for event time series

Researcher

Ryota Kobayashi

Ryota Kobayashi

Research Organization of Information and Systems
National Institute of Informatics
Assistant Professor

Outline

Though the demand for utilizing the time series data is growing, time series analysis has not been widespread due to the difficulty in developing a time series model. The experts have to construct a “good” time series model, because the performance critically depends on the model. In this project, we focus on the event time series that is the time stamp of event occurrence, and we develop a machine-learning framework that can analyze the event time series in an automatic way.

Edgar Simo-Serra

Interactive AI-Aided Content Creation using Deep Unsupervised Learning

Researcher

Edgar Simo-Serra

Edgar Simo-Serra

Waseda University
School of Fundamental Science and Engineering
Junior Researcher

Outline

This research focuses on assisting content creation with artificial intelligence with a focus on illustrations. The traditional illustration process both requires expert knowledge as well as lots of effort. However, by training neural networks with large amounts of illustration, this research aims at creating tools that allow amateurs to reproduce the abilities of professionals. Although initially aimed at illustration, potential applications include natural language processing and image processing.

Kumiko Suzuki

Multimodal learning-based forest structure estimation using earth observation data

Researcher

Kumiko Suzuki

Kumiko Suzuki

KOKUSAI KOGYO CO.,LTD.
R&D Department
Research Officer

Outline

This research focuses on assisting content creation with artificial intelligence with a focus on illustrations. The traditional illustration process both requires expert knowledge as well as lots of effort. However, by training neural networks with large amounts of illustration, this research aims at creating tools that allow amateurs to reproduce the abilities of professionals. Although initially aimed at illustration, potential applications include natural language processing and image processing.

Maina Sogabe

Break the Limit of Live Imaging Observation by Sparse Modeling

Researcher

Maina Sogabe

Maina Sogabe

Kyoto University
Institute for Frontier Life and Medical Sciences, Graduate school of medicine
Ph.D. Course Student

Outline

Live imaging is a novel method which is received the most attention in the field of life science and medicine. However, detailed observation of biological events in the live imaging for a long time often suffers from the fluorescence diminishment. To solve the crucial problem in the live imaging, my project aims at developing a new technique for obtaining the detailed cell morphological data, by using the sparse modeling, which can elucidate the most relevant part from a small amount of data, and by the reduction of the laser irradiation mass for decrease in the damage to fluorescence protein.

Shinji Takaki

Text-to-speech synthesis framework for arbitrary speech

Researcher

Shinji Takaki

Shinji Takaki

Research Organization of Information and Systems
National Institute of Informatics
Project Assistant Professor

Outline

Recently, the performance of text-to-speech (TTS) systems has been significantly improved so that using them is getting popular in various speech applications. However, TTS systems could be used in limited situations because narration like speech is mainly generated from them. Also, there are various types of speech which human can utter but TTS systems can’t generate. In this work I propose a novel text-to-speech framework in which arbitrary speech can be generated.

Naoya Chiba

3D Measurements Based on Estimation Light Tranport Matrix between Camera and Projector

Researcher

Naoya Chiba

Naoya Chiba

Tohoku University
Graduate School of Information Sciences
Master Course Student

Outline

There are various 3D measurement methods based on the use of a projector and a camera have been proposed. However, there is still difficulty measuring metallic objects and translucent objects. We use Light Transport Matrix, that is able to describe optical properties of measurement objects, as an optical model for adaptive measurement in a variety of scenes. To achieve efficient measuring of Light Transport matrix, we use feedback to generate projection patterns.

Zheng Yinqiang

Developing a Multispectral RGB-D Camera for 3D Video Capture of Underwater Scenes

Researcher

Zheng Yinqiang

Zheng Yinqiang

Research Organization of Information and Systems
National Institute of Informatics
Assistant Professor

Outline

The objective of this research is to develop a novel multispectral RGB-D camera for 3D video capture of underwater scenes. In contrast to existing 3D surface shape recovery methods and devices, the proposed camera utilizes the light absorption property of water, and estimates the light travelling path in water on the basis of the absorption difference at two close but different wavelengths. It is expected that the camera can be used to capture the RGB color and 3D shape of fast moving objects in water.

Kei Terayama

Development of technology for bluefin tuna in fish farm with camera and imaging sonar

Researcher

Kei Terayama

Kei Terayama

The University of Tokyo
Graduate School of Frontier Sciences
Post-Doc Researcher

Outline

Supply of tuna, especially pacific bluefin tuna, is now considered to be reaching the point of depletion as they has been overfished in recent years. To supply pacific bluefin tuna stably, improvement of fully closed life-cycle aquaculture is essential. In this project, we will record movies of tuna fry (young tuna) with both imaging sonars and cameras in a fish farm, and develop a behavior monitoring system for them based on computer vision and machine learning techniques. The goal of this project is to reveal the death of fry that is the bottleneck in tuna farming and to improve survival rate of tuna.

Kazutaka Nakashima

Actual fabrication aware geometry optimization

Researcher

Kazutaka Nakashima

Kazutaka Nakashima

The University of Tokyo
Graduate School of information Science and Technology
Ph.D Student

Outline

Recently, a lot of innovative technologies for fabrication (e.g., 3D printing) emerged. Accordingly, in the field of computer graphics, many researches about fabrication are presented. However, most of them only optimize a shape and give it some functionality, and ignore cost for the actual fabrication. In my research, I also take such cost into consideration and optimize a whole of the fabrication process.

Yukino Baba

Machine learning for reliable peer-assessment systems

Researcher

Yukino Baba

Yukino Baba

Kyoto University
Graduate School of Informatics
Assistant Professor

Outline

On-demand recruiting systems have been applied in a variety of businesses, such as taxi, baby sitting, or translation. Due to a large number of constructors involved in on-demand-recruiting, conventional approaches for constructor assessment are hardly to be applied to. Thus, peer assessment is incorporated into on-demand-recruiting systems as an alternative way; however, we cannot assume that all the given evaluations are reliable. In this study, given the information on the peer assessment systems, I will develop machine learning methods for estimating constructor assessment in a reliable manner.

Yuko Hara-Azumi

Machine learning for reliable peer-assessment systems

Researcher

Yuko Hara-Azumi

Yuko Hara-Azumi

Tokyo Institute of Technology
School of Engineering
Associate Professor

Outline

This proposal will realize a novel tiny processor that achieves performance-and-energy-efficiency and flexible customizability for a variety of embedded and Internet-of-Things (IoT) applications and build its design automation environment. The proposed processor enables real-time processing of streaming data that is produced continuously in the Big Data era, only within the IoT edge devices, leading to creation of novel businesses/services and solutions of social issues.

Yuki Funabora

Basic Researches of Structure and Control System for Realizing Innovative Flexible Wearable Assist System

Researcher

Yuki Funabora

Yuki Funabora

Nagoya University
Graduate School of Engineering
Assistant Professor

Outline

For realizing an Innovative Flexible Wearable Assist System (IFWAS) consisting enormous combination of thin artificial muscles, this research clears both a basic topology of thin artificial muscles and a novel control method suitable for that. IFWAS has possibility of assisting user’s flexible motions in daily life. Minimal spatial combination of artificial muscles (minimal module) will be cleared in this research term, being able to express basic motions of human body parts (flexion-extension, abduction-adduction, lateral-medial rotation). Also a novel control method will be developed in order to control the module to realize any three dimensional motions.

Yusuke Matsui

Quantized Linear Algebra: Fast and Memory-efficient Matrix Computation by Data Quantization

Researcher

Yusuke Matsui

Yusuke Matsui

Research Organization of Information and Systems
National Institute of Informatics
Researcher

Outline

We propose a fast and memory-efficient matrix computation scheme, Quantized Linear Algebra (QLA). QLA first compresses input vectors or matrices into short memory efficient codes. The mathematical operations over the codes such as the matrix product are efficiently computed using looking up techniques. By the proposed QLA, computationally heavy operations including large scale machine learning can be achieved using small scale computational resources.

Tomoyuki Morimae

Secure cloud quantum computing with classical verifier

Researcher

Tomoyuki Morimae

Tomoyuki Morimae

Kyoto University
Yukawa Institute for Theorecitacl Physics
Lecturer

Outline

In this proposal, I plan to develop a new protocol of secure cloud quantum computing where a completely classical verifier can delegate quantum computnig to a remote sever without leaking any privacy.

Naoto Yanai

Universal Vulnerabilty Testing Method Based on Cryptographic Expection Distribution for Source Code Level

Researcher

Naoto Yanai

Naoto Yanai

Osaka University
Graduate School of Information Science and Technology
Assitant Professor

Outline

Vulnerabilities are important flaws of security for an information system, and the existing analysis against the vulnerabilities focuses on an individual testing including each attack. However, due to requireing a large number of verifications, such an approach brings on a downgradation of the producivity and a potential weakness depending on a lack of the executions. In this research, we formalize vulnerabilities based on expection distributions of programs from a standpoint of cryptography. Next, for constructing program codes, we propose a method for vulnerability testing to guarantee universal security which is independent of the individual testing described above.

Yutaro Yamaguchi

Exploring the Limitation of Matroidal Properties in Packing Problems on Graphs

Researcher

Yutaro Yamaguchi

Yutaro Yamaguchi

Osaka University
Graduate School of Information Science and Technology
Assistant Professor

Outline

A packing problem on graphs is to find disjoint structures in a given graph. A variety of packing problems are known to be tractable, where matroidal properties enjoyed by the solution sets play important roles. This study aims to explore the boundary of such matroidal frameworks, and to provide a characterization of tractable packing problems in terms of matroidal properties.

Kazuhiko Yamamoto

An Optimization of 3D Audio Human Related Transfer Function for A Specific User using Machine Learning

Researcher

Kazuhiko Yamamoto

Kazuhiko Yamamoto

Yamaha Corporation
Research Division
Senior Researcher

Outline

To realize 3d spatial sound rendering in a virtual environment with two channels headphone, Head Related Transfer Functions (HRTF) is required. This HRTF highly depends on an individual property of a human. Because of this, it must be required to measure it for each user respectively. However, it is difficult due to the tedious and expensive measurement process requiring an anechoic chamber. To address this, we propose a fully perceptual-based HRTF fitting method to a specific user using machine learning technique.

Ryo Yonetani

Privacy-Preserving First-Person Vision

Researcher

Ryo Yonetani

Ryo Yonetani

The University of Tokyo
Institute of Industrial Science
Research Associate

Outline

First-person vision is one of the emerging topics in computer vision, which makes use of wearable cameras for various recognition tasks. In this work, we aim to develop first-person vision techniques that recognize everyday activities while preserving the privacy of camera wearers and that of people who come into the view of a camera.

Shoko Wakamiya

A Study on Medical Applications of Social Media by Solving Spatio-temporal and Semantic Gaps

Researcher

Shoko Wakamiya

Shoko Wakamiya

Nara Institute of Science and Technology
Institute for Research Initiatives
Postdoctoral Researcher

Outline

The power of social media has not been efficiently utilized in medical fields, though lots of tweets concerning people’s health condition or diseases are shared over social media like Twitter. Several diseases such as hay fever can be observed by regarding Twitter users as social sensors and exploiting their tweets. In this study, we aim at developing a Twitter-based surveillance system for hay fever by solving spatio-temporal and semantic gaps that are important issues of social sensors.

Program

  • CREST
  • PRESTO
  • ACT-I
  • ERATO
  • ACT-X
  • ACCEL
  • ALCA
  • RISTEX
  • AIP Network Lab
  • Global Activities
  • Diversity
  • SDGs
  • OSpolicy
  • Yuugu
  • Questions