[Information and future] Acceleration Phase Year Started: 2019

Affiliation and job title should automatically appear from the information that a researcher registered with researchmap.
Data may be outdated or undocumented.
When there is not a connection via the internet, data are not displayed.

Yoshitaka Ushiku

Creation of technology to automatically generate captions to various data


For image / video caption generation, which is one of the ultimate forms of media understanding, there are fundamental requirements, (i) correspondence to individual users, (ii) correspondence to detailed expressions, (iii) data without ground truth captions, and (iv) flexibility for varied stories. Although it is inherently necessary to deal with it, it is still unattainable. In this research, we integrate advanced knowledge in various fields such as natural language processing, image recognition and machine learning, realize these functions, and enable caption generation from various kinds of data.

Shuhei Kurita

Knowledge extraction from large corpora with deep neural network and semantic analysis


We propose new neural network-based models that learn the general knowledge and activate them. Especially, we study distributional representations of knowledge learned from external resources such as Wikipedia. We have already developed semantic analysis models with deep reinforcement learning approaches in the previous term. We apply them and attempt to construct pre-trained neural network models with the knowledge that can be applied for general purposes. We also investigate a new system that can be trained via textual directions using deep reinforcement learning. This is for the new knowledge activation.

Tsutomu Kobayashi

Assistance of strengthening assertions for program codes through consistent abstraction


Various kinds of conditions such as invariants, preconditions, and postconditions can be considered for a program code. Adding specifications on such conditions (assertions) to code and using it for generating tests and debugging is known to be effective for improving the quality of program code. However, manually specifying strong assertions is hard for developers and automatic methods for generating assertions have limitations on assertion strength. In this research, we aim to generate formal specifications of a given program code through abstraction and do analyses on formal specifications to strengthen assertions of the given code.

Shigeyuki Sato

Domain-specific compilers based on parallel patterns


Such computation as to apply local computation regularly to input data frequently appears in scientific computing and machine learning. Describing it with parallel patterns, which abstract local computation, brings the versatility and reusability of programs, and enhances productivity. This study aims to obtain mechanically on the basis of domain- or pattern-specific optimizations as efficient implementations of such highly abstract programs as proficients do.

Hiroaki Shiokawa

Time and Space Efficient Big Data Processing via Data Skewness Caching


Large-scale data analysis has been successful in a wide range of fields such as business, medical care and so on. However, a high-performance computer is indispensable so as to analyze such large-scale data quickly. The goal of this project is to establish time and space efficient algorithms that can handle large-scale data on the limited computation resources. The key idea of this project is to design data processing algorithms as to reduce redundant computations by using frequent data structures hidden in the data. With our proposal, we aim to improve feasibility of the large-scale data analytics.

Chenhui Chu

The Semantic Typology of Visually Grounded Paraphrases


Visually grounded paraphrases (VGPs) that we propose originally, are different phrasal expressions describing the same visual concept in an image. Our research until now treats VGP identification as a binary classification task, which ignores various phenomena behind VGPs. In this project, we aim to create semantic typology of VGPs to elucidate the phenomena behind VGPs and open up novel ways of utilizing VPGs for various language and vision tasks, which require semantic understanding.

Shuichi Hirahara

Analyzing the Complexity of Minimum Circuit Size Problem Toward Establishing Secure Cryptography


The Minimum Circuit Size Problem is one of the central problems in computational complexity theory. The problem asks for searching a minimum computer hardware, and thus it is a natural and fundamental problem. Recently, some relationship between this problem and cryptography was found, but its computational hardness remains elusive. This project aims at analyzing the computational complexity of the Minimum Circuit Size Problem.

Takashi Miyamoto

Earthquake damage detecion by machine learning of spatiotemporal big data obtained from sattelite remote sensing


At the time of large earthquake occurrence, damage distribution information on where and what kind of damage occurred is indispensable. In this research, we will construct a machine learning method for spatiotemporal big data, which detects the damage distribution of each house from multi-temporal satellite imagery. We will try to develop an optimal machine learning model considering the characteristics of satellite imagery data such as data-imbalance and periodicity.

Shogo Yamashita

Developing a 3D Water Flow Measurement Technology for Swimming Pools


Understanding water flow around a swimmer is key to reducing the water resistance when swimming and improving propulsive force. However, existing water flow measurement technologies are not suitable for measuring human swimmers because they can only measure a limited area and have potential adverse health effects on swimmers. In this research, we propose a harmless method of water flow measurement using food-grade particles and a harmless light source. Proposed water flow measurement technology for swimmers should be able to contribute toward creating a more efficient swimming form, which remains unexplained thus far.

Hironori Yoshida

Upcycling Tree Branches as Furniture and Architectural Elements through Digital Design and Fabrication


Diverse natural materials such as stones and woods have been used as architectural elements preserving their native forms since primitive shelters, however, the use of them in modern buildings is limited due to their irregular properties. In this project, we take the diversity as a playful input for design, and develop a human-in-the-loop workflow for design and fabrication of architectural elements. Taking tree branches as a material in native forms, we aim to fabricate networks of branches and realize it by CNC milling process. Each connection has a customized unique joint adapted to the native forms of branches. Together with low-cost mobile scanning devices and collaborative interface between users and digital fabrication tools, users can easily contribute to the process of making, which was an advantage of traditional architecture with local materials.

Lijun Liu

Development of a method for high-performance multiscale analysis with time-space parallel computing


Characteristic prediction by high precision calculation method is critical for accurate evaluation and new development of materials and structures. In this research, we aim to establish a multi-scale analysis method using temporal and spatial parallel computations that realize long time simulation. Through the application of the proposed method in such as composition analysis of steel materials and characterization of electronic devices, we verify the effectiveness of this simulator and challenge to create simulators that can be widely used.

Quick Access


  • ACT-I
  • ACT-X
  • ALCA
  • AIP Network Lab
  • JST ProjectDB
  • Global Activities
  • Diversity
  • SDGs
  • OSpolicy
  • Yuugu
  • Questions