| 小池 和正 | 1351041: M, 1回目発表 | コンピューティング・アーキテクチャ |
|
title: A Proposal for partial redundancy technique of the many-core architecture
abstract: Recently,The many-core architecture that improve parallelism by implementing many processing-core ,aiming to improve the entire performance is paid attention.On the other hand, the increase of soft errors caused by the miniaturization of the manufacturing process is seen as a problem in the architecture. Method to execute redundantly each instruction is useful as a countermeasure to this problem, but overhead is increased.In this presentation, the aim is to reduce overhead and to remain high reliability by duplicating the important part in the program,I explain partial redundancy method based on the importance of the instruction. language of the presentation: Japanese 発表題目: メニコア・アーキテクチャにおける部分的冗長化手法の提案 発表概要: 近年，大量のプロセッサ・コアを搭載することにより並列性を向上し，性能向上を目指すメニコア・アーキテクチャが注目されている．しかし一方で，プロセスの微細化に伴うソフトエラーの増加が同アーキテクチャにおいても問題視されている．この問題への対策としては各命令を冗長的に実行する手法が有用であるが，冗長化のためのオーバーヘッドが大きくなるのが課題である．本発表では，重要な部分のみ冗長化を行うことにより，高信頼性を保ったままオーバーヘッドを削減することを目的とし，命令の重要度に基づいた部分的冗長化手法について述べる． | ||
| 尾上 紗野 | 1351025: M, 1回目発表 | ソフトウェア工学 |
|
title: A Study of the Characteristics of Developers' Activities in GitHub
abstract: What types of developers do active software projects have? This paper presents a study of the characteristics of developers' activities in open source software development. GitHub, a widely-used hosting service for software development projects, provides APIs for collecting various kinds of GitHub data. To clarify the characteristics of developers' activities, we used these APIs to investigate GitHub events generated by each developer. Using this information, we categorized developers based on measures such as whether they prefer communication by coding or comments, or whether they are specialists or generalists. Our study indicates that active software projects have various kinds of developers characterized by different types of development activities. language of the presentation: Japanese | ||
| 米田 純 | 1351043: M, 1回目発表 | ユビキタスコンピューティングシステム |
|
title: Appliance control system with indoor localization system
abstract: A remote control is commonly used for controlling a appliance in our home. However, the increase of appliances causes the number of remote controls. As a purpose of reducing remote controls, there exists multifunctional learning remote controls. But its too many buttons may confuse seniors and beginners. On the other hand, gesture based control system have been recently proposed. However, it require users to perform complicated gestures for selecting appliance or changing channels. Therefore, it is bothering and not practical. In our research, we propose the appliance control system having both advantage of remote controls and gesture recognition. A user selects the target device only pointing the appliance with a remote control. To achieve this, we utilize the high-accuracy indoor localization system. After selecting the device, a user can control the appliance with the buttons displayed on the remote control according to the selected device. Especially, we focus on how to estimate the desired appliance from the direction of the remote control accurately. language of the presentation: Japanese 発表題目: 屋内位置情報を利用した家電操作システムの提案 発表概要: 一般家庭において、家電操作はそれぞれの家電に付属のリモコンにおいて行われることが多いが、家電の種類が増えるにつれ、リモコンを探したり、選ぶといった煩雑さが多くなり問題となっている。この問題に対して、１台 のリモコンで複数の家電を操作できる学習リモコンが発売されている、しかし家電の選定をボタン操作で行う点や、ボタンの多さから、操作の煩雑さの十分な解決には至っていないと考えられる。また、直感的な入力方法として、ジェスチャ認識による家電操作法が挙げられるが、ボタン操作のように素早く複雑な操作を行うことは難しい。そこで本研究では、ジェスチャ認識とリモコンの長所を兼ね備えた新しい家電制御システムを提案する。提案システムは、高精度に屋内位置情報が得られることを前提とし、リモコンで家電を指すという直感的な動作で制御対象の家電を選択する。その後は、選択された家電に応じて液晶に表示されるリモコンを用いて、ボタン操作により家電を制御する。本研究では、特にリモコンの向きを用いた所望家電の推定アルゴリズムを提案し、その精度改善に取り組む。 | ||
| 尾上 栄浩 | 1351026: M, 1回目発表 | 計算メカニズム学 |
|
title: XML tree structure compression using RePair[Paper Introduction]
abstract: XML is one of the most widely-used formats today.However, due to the repetitiveness of XML markup, XML documents tend to include redundancy. To solve the problem, many compression methods for XML documents have been already proposed. In particular, SLCF tree grammars are known to be efficient memory representations of XML document trees, over which we can traverse without decompression. In this presentation, I will introduce TreeRePair algorithm to compress tree structure using RePair algorithm. TreeRePair is a linear time algorithm for computing a small SLCF tree grammar that generates a given XML document tree. TreeRePair algorithm consists of two phases, namely the replacement phase and the pruning phase. During the replacement phase, a most frequent digram is subsequently replaced by a new nonterminal. And during the pruning phase, unprofitable nonterminals which deteriorate compressibility are eliminated. By experimental results, it was demonstrated that TreeRePair greatly outperforms other SLCF tree grammar based compressors. language of the presentation: Japanese 発表題目: RePairアルゴリズムを用いた木構造の圧縮[論文紹介] 発表概要: XMLはその汎用性の高さから幅広く用いられているが，XML文書は冗長でサイズが大きくなりやすいという問題点を持つ．そのため，XML文書の圧縮法がいくつも提案されている．特に，SLCF木文法を用いた圧縮法においては，圧縮されたXML文書に対して，圧縮されたままの状態で効率的な走査が可能であることが知られている．このような性質はリソースの限られた環境において非常に重要となる．本発表では，RePairアルゴリズムを用いて木構造を圧縮するTreeRePairアルゴリズムについて紹介する．TreeRePairは，与えられたXML文書木のみを生成するサイズ最小のSLCF木文法を求める，線形時間の準最適アルゴリズムである．TreeRePairはreplacement phaseとpruning phaseの2つの phaseから構成される．replacement phaseでは最頻出のdigramを順に非終端記号に置き換え，pruningphaseでは結果として圧縮率を悪化させる非終端記号を除去する．実験により，TreeRePairはSLCF木文法に基づく他手法より優れた圧縮率を達成することが示されている． | ||
| 立古 佳大 | 1351112: M, 1回目発表 | 情報基盤システム学 |
|
title: Introduction about power save methods of wireless communication abstract: Energy problem of wireless communication As mobile devices become common, energy consumption of wireless communication become big issue. In this presentation, I will explain several power saving methods and problems of them. language of the presentation: Japanese | ||
| 田ノ元 正和 | 1351066: M, 1回目発表 | コンピューティング・アーキテクチャ |
|
title:An Analysis of Performance Overhead in a Global Atmospheric Model
abstract: Recently, scientific computing is expected more precision and execution speed with computer performance growing up. However, a Global Shallow Water Atmospheric Model for climate change simulation is achieved only 8% execution speed against peak performance in Intel MIC execution. In this presentation, I analyze this performance overhead and discuss about how to improving performance. language of the presentation: Japanese 発表題目: 全球大気モデルシミュレーションにおける性能低下要因の解析 発表概要: 近年のコンピュータの性能向上に伴って、科学技術計算においても従来に比べ精度や実行速度の向上が期待されている。 しかし気候変動のシミュレーションに用いられるShallow Water Equation(SWE)を用いた全球大気モデルでは、Intel MIC上においてノードあたりの実行速度がピーク性能比8％と十分な性能を得られていない。 本発表では、この性能低下の要因について解析し、さらに改善に向けた検討を行う。 | ||
| KRICHENBAUER MAXIMILIAN MICHAEL | 1351121: M, 1回目発表 | インタラクティブメディア設計学 |
|
title: Augmented Reality User Interface for Media Production
abstract: Current media productions in all areas of film, TV and games make extensive use of 3D CG content. However, producing and handling this virtual 3D content is commonly performed with 2D user interfaces similar to those of office work, which makes the work very indirect and unintuitive and may result in reduced quality and efficiency. We propose the possibility of using Augmented Reality (AR) and real-time hand tracking to create virtual 3D user interfaces that allow the artists to interact with their content very much as they would with a physical medium. Research is proposed as to how the special requirements of professional media production can be implemented in 3D user interfaces in general. language of the presentation: English | ||
| PADERNA RYAN R. | 1351123: M, 1回目発表 | ネットワークシステム学 |
|
title: Modified Matching Pursuit based Channel Estimation for ISDB-T abstract: This research demonstrates a new approach in reliable channel estimation for Integrated Services Digital Broadcasting Terrestrial (ISDB-T). The assumption is that the channel characteristics of the succeeding symbols doesn't change much. With this assumption. the computational cost for estimating the channel will be reduced using Modified Matching Pursuit Algorithm (MMP). The result of the simulation shows that the MMP was able to validate decent channel estimation with reduced computational cost. language of the presentation: English | ||
| 橋本 稜平 | 1351086: M, 1回目発表 | 環境知能学 |
|
title:High Performance Control for Graceful Motion of an Intelligent Wheelchair
abstract: To be acceptable to human drivers, the motion of an intelligent robotic wheelchair must be more than just collision-free: it must be graceful. We define graceful motion as being safe, comfortable, fast and intuitive. we quantify these properties of graceful motion, providing formal evaluation criteria. We propose a method for graceful motion and present implementation results for the task of driving through a narrow doorway, evaluated on a simulated model of the wheelchair. Existing methods typically respond to tight clearances with very slow motion which is not graceful. Our results show that, starting from a set of representative poses, the wheelchair passes through the door at near maximum speed, staying close to the mid-line of the doorway. The velocity of the wheelchair reflects the curvature of the path rather than the closeness of the door edges, so it can move smoothly, safely, and quickly through the doorway. Thus, this study makes two contributions - first it introduces the concept of graceful motion and provides quantitative measures for the same, and second, it proposes a method for graceful motion and demonstrates it on a specific task. language of the presentation: Japanese | ||
| 大谷 まゆ | 1351018: M, 1回目発表 | 視覚情報メディア |
|
title: [Paper Introduction] Videoscapes: Exploring Sparse, Unstructured Video Collecitons
abstract: With a growing number of videos that are shared through the internet, an efficient browsing system is demanded. Some systems for photo collections have been proposed but are not extended for video collections. This talk explains a novel browsing system called videoscape which deals with an unstructured video collection. To represent the video collection, videoscape uses a graph structure whose edges denote videos and nodes denote a set of frames in which a user can move to another video seamlessly. Videoscape also provides an interface that allows users to brows multiple videos without losing their spatial positions in the scene based on a three-dimensional reconstruction technique. Introducing this system and its limitation, I summarize the extension of videoscape toward event summarization which considers semantics of videos. language of the presentation: 日本語 発表題目: [論文紹介]Videoscapes: Exploring Sparse, Unstructured Video Collecitons 発表概要: webサービスの発展により，大量の動画が web上で共有されるようになった.このような大量の動画を効率よく視聴するための手法が要求されている．従来，写真コレクションを閲覧するための手法が提案されてきたが，それらが動画に用られることは少なかった．本発表では動画コレクションに対して適用可能な手法としてVideoscapeを紹介する．Videoscapeはエッジが動画を，ノードが動画中の遷移可能フレームを表すグラフによって動画コレクションを表現する．遷移可能は動画同士には遷移フレームに共通に表れるランドマークがあり，視点移動を抑えたトランジションが可能になっている．本発表ではVideoscapeとその課題をまとめ，最後に動画を用いたイベント要約手法の展望を述べる. | ||
| 大内 啓樹 | 1351014: M, 1回目発表 | 自然言語処理学 |
|
title: Effect of Conjunctive Expression Recognition for Dependency Parsing and the Future Direction
abstract: Dependency parsing is the task of identifying the dependency relations between words. The accuracy of dependency parsing exceeds 90%, and this has been applied to various applications, such as summarization, translation and information retrieval. However, there is still room for improvement on analysis of complex sentences that include subordinate or embedded clauses in dependency parsing. A means of improving it is to discriminatine the part of speech (POS) tags of subordinate conjunction and others, because they are annotated the same POS tag. We distinguish subordinate conjunction and other POS, and verify that the discrimination helps to increase the accuracy of English dependency parsing. In this presentation, we explain the experiment in more detail and describe the future direction of our research. language of the presentation: Japanese 発表題目: 依存構造解析における従属接続詞認識の効果と今後の展望 発表概要: 単語間の係り受け関係を同定するタスクである依存構造解析は，解析精度が90%を超え，文書要約や翻訳，情報検索など様々なものに応用されている．しかし，従属接続詞や埋め込み節を含む複雑な文の解析にはまだ改善の余地がある．1つの改善策として、従属接続詞と前置詞といった異なる品詞に同じ品詞タグが付与されているので、それらを分別することが考えられる。我々は、従属接続詞と他の品詞を分別し、異なる品詞タグを付与することよって依存構造解析精度が向上することを確認した．本発表では，実験の詳細と今後の研究の方針を説明する． | ||
| 島田 健史 | 1351055: M, 1回目発表 | ロボティクス |
|
title: Object and Action Recognition in 3D using First-Person Vision
abstract: There are several attempts that use imitation learning to effectively teach a robot motions and skills of tasks. To achieve such attempts, recognition and measurement of human activity are required. In this research, we focus on a method for measuring and recognizing human activity by first-person vision, which consists of an eye tracking system and an RGB-D camera. In this presentation, we would like to introduce technique to detect moving objects from freely moving camera and technique to recognize human activity. Finally, I would like to explain our current research progress. language of the presentation: Japanese | ||
| 吉本 公則 | 1351111: M, 1回目発表 | ロボティクス |
|
title:
Object-Relationship Description Based on the Topological Information
abstract: Describing the relationship between objects is important for a robot to understand the meaning of the task. The relationship is usually described by contact information. However, one cannot recognize the task where no or less contacts happen from such description. To solve this issue, we focus on the method that uses the topological information to describe the relationship. Although the prior method can only handle with the topological information between two curved lines, we discuss to improve the method to handle with line-surface and two-surface relationship. language of the presentation: Japanese | ||
| 村岡 伸哉 | 1351107: M, 1回目発表 | 計算システムズ生物学 |
|
title: Development of dynamic models of proteins for molecular animation
abstract: A method to visualizing the protein molecule reactions in our body is to make animation. In conventional animation, motion of proteins in our body has not been considered. In this study, using the results of the normal vibration analysis that simulates the fluctuation of the protein molecule, I introduce a method for adding dynamic motion to protein models and fit it to those. language of the presentation: Japanese 発表題目 : 分子アニメーション制作のための動的なタンパク質分子モデルの開発 発表概要 : 生体内のタンパク質分子が関与する反応を表現する方法の１つとして、アニメーションを作製するという方法がある。従来のアニメーションでは、タンパク質分子の生体内での動きが考慮されていなかった。そこで本研究では、生体内でのタンパク質分子の揺らぎをシミュレーションした基準振動解析の結果を用いて、アニメーションで使用されるタンパク質分子モデルに、生体内で予想される動的な動きを与える手法の提案と実装を行う。 | ||
| HOLLAND MATTHEW JAMES | 1351119: M, 1回目発表 | 数理情報学 |
|
title: Site-adaptive probabilistic wind speed forecasting under data constraints abstract: To ensure the long-term economic viability of wind-based energy projects, the intermittent nature of the wind must be addressed at an operational level. Anticipatory control systems and non-linear turbine controllers have been proposed as effective means to reduce costs and improve output efficiency, but they require reliable forecasts of local, short-term future wind speeds. Traditional statistical approaches generally focus on a specific site or region, and make modeling assumptions based on prior meteorological analysis. In general however, access to the data necessary to arrive at such insights may be constrained, and a more site-adaptive model becomes required. We propose a forecasting model under the continuous ranked probability score-minimizing framework using a Weibull based approach, as well as an intuitive model-building feature selection algorithm. Having derived a closed-form expression for this optimization metric, we carry out tests on the database of AMeDAS observations. Our proposed model is shown to outperform deterministic and stochastic references at forecast lags beyond two hours. language of the presentation: Japanese | ||
| 大津 淳 | 1351019: M, 1回目発表 | ロボティクス |
|
title: On the necessary factors to generate human-like motions for an android abstract: I aim to generate human-like motion for an android robot and thus would like to clarify the necessary factors in order to achieve this aim. At the first step, I would like to use computer graphics characters with various figures, such as a realistic android robot, a humanoid robot, and a simplified model used lines. I assume that the motion obtained by a motion capture system is human-like one and I propose to use a subjective evaluation by comparing the original and the altered motions. I would like to introduce two related works: one focuses on alteration of the motion and the other focuses on the relation between motion naturalness and environmental factors. language of the presentation: Japanese | ||