研究発表

This poster provides an interim report on our MSRA CORE 14 project.

Astrophysics can be regarded as the ultimate remote sensing because no one can travel to see real heavenly bodies located far from the Earth. The only option for astronomers is to observe and study a variety of signals emanating from the distant bodies. This might lead to their enthusiasm for data visualization, compared to ones in other disciplines. Indeed, astrophysics relies heavily on the values of observed data visualization and analysis. In this talk, an overview of the latest research results from our collaborative research projects with astronomers will be given, including asymmetric biclustering of multivariate data for correlated subspace mining and its application to supernovae categorization, TimeTubes for visually extracting feature polarization variations from long-sequences of observed blazar datasets, and aflak as a novel visual programming environment to conduct fine-grained transformations, filtering and visual analyses on multispectral astrophysical observations.

SIGGRAPH2018の参加報告を中心とした最新CG技術の研究開発動向の紹介．

In this lecture, after providing a brief history and basic concepts, the power and values of AR/VR visualization will be illustrated with several results from the related research projects which the lecturer has conducted so far. These examples include lidar measurement and detection of wake turbulence at an airport, deploying an infinite number of pressure meters in a hybrid wind tunnel, haptization of 3D Diffusion Tensor-MRI datasets, and synesthesia/optic illusion-based enhancements of multisensory/visual display.

今月カナダBIRSで開催されたRestructuring IEEE VIS for the Future国際ワークショップにおける集中的討論の内容をベースに，今後のデータ可視化技術の研究開発および応用の方向性について，特に没入的表示や人工知能技術の利活用に焦点をあてて紹介した．

Sparse modeling is a general principle for promoting high-dimensional data-driven science. I am head of a planned visualization research project in a designated governmental program in Japan. The primary goal of this research project is to construct a visualization platform for understanding the behavior of given high-dimensional measurement data in physical space by projecting it onto a judiciously designed information space. In this talk, I will give an overview of the governmental program and introduce the latest research results from the research project, including spectral-based contractible parallel coordinate plots, biclustering multivariate data for correlated subspace mining, and TimeTubes for visualizing polarization variations in blazars in a personalized immersive display environment.

同じ分野に30余年も関わっていれば，ひとつやふたつは話のネタになりそうなことにも出会すものである．本講演では，主として若手の研究者や学生さんに向けて，自身の研究に繋がったエピソードを三例ほど紹介してみたい．「いつかは使ってやろう！」（1995年），「眺めることのススメ」（2005年），「素材は日常至るところに」（2015年）．

SIGGRAPH2017の参加報告を中心とした最新CG技術の研究開発動向の紹介．

TimeTubes is a visualization tool for observed datasets of blazars. It allows astronomers to geometrically analyze the dynamic behavior of and relationship among multiple time-varying variables in the datasets. This poster mainly focuses on a novel functionality of TimeTubes—visual data fusion, which is intended to visually fuse multiple datasets from different observatories for the same blazar.

Innovative Areas called “Initiative for High-Dimensional Data-Driven Science through Deepening of Sparse Modeling” (http://sparse-modeling.jp/index_e.html). In this project, I have been heading a planned visualization research project with a primary goal to construct a visualization platform for understanding the behavior of given high-dimensional data in physical space by projecting it onto a designated information space. For this purpose, we transform relatively low-dimensional data obtained through sparse modeling to two, three, or four-dimensional space where we can fully visualize the characteristics of the original high-dimensional data. The idea behind our project is to establish an interactive model called human-in-the-loop by incorporating visual feedback from the users in the analysis of high-dimensional data based on sparse modeling. In this talk, I will give an overview of the governmental project and introduce recent research results from the planned research project, including spectral-based contractible parallel coordinate plots, Biclustering multivariate data for correlated subspace mining, and TimeTubes for visualizing polarization variations in blazars.

通常の投影型ディスプレイを用いてボリュームデータを可視化するには，フィールド値に色や反騰明度を割り当てて投影変換するしかない．その際に生じるモデリングやシミュレーション・計測データ解釈の不確実さをできる限り取り除き，重要な特徴を解析者に注目させ，また誘目するための離散微分位相幾何学アプローチの概要を紹介した．

本ポスタでは，同領域計画研究で開発中の多対多平行座標系表示と時系列多変量可視化TimeTubesの進捗状況に関して，可視化の「一覧性」の向上という観点から紹介した．

During FY2013-2018, we have a governmental project under a Grant-in-Aid for Scientific Research on Innovative Areas called “Initiative for High-Dimensional Data-Driven Science through Deepening of Sparse Modeling” (http://sparse-modeling.jp/index_e.html). In this project, I have been heading a planned visualization research project with a primary goal to construct a visualization platform for understanding the behavior of given high-dimensional data in physical space by projecting it onto a designated information space. In this talk, I gave an overview of the governmental project and introduce recent research results from the planned research project, including spectral-based contractible parallel coordinate plots, Biclustering multivariate data for correlated subspace mining, and TimeTubes for visualizing polarization variations in blazars.

本稿では，スパースモデリングによって得られた比較的次元が低いデータを，具体的に効果的な視覚解析に落とし込むいくつかの事例について紹介した．ここで用いる道具は，ある対象間の関係などを表す抽象データの視覚解析の表現としてよく用いられるネットワーク(グラフ) である．実際ネットワーク表現は，ソーシャルネットワークや同時に購入する物品関係などの抽象データを効果的に可視化することができ，人によるデータの視覚解析を支援し促進することができる．本稿では，多変量数値データに変換操作を施してネットワーク表現することで，人の視覚解析を効率化できたいくつかの成果事例について紹介した．

コンピュータの能力が進歩し，人間の脳を超える日が来るかもしれない，と言わ れていますが，本当にそうでしょうか？本ラウンドテーブルでは，コンピュータと協調する人間の脳の振る舞いに焦点を当てて議論を展開した．藤代は，錯視を利用した3次元ポップアップディスプレイの研究開発を事例にして参加した．

開発中の専用マークアップ言語LMMLの処理系を用いて，法科学に関連するビッグデータをどのように統合的に記述し，視覚分析して，犯罪捜査や裁判の電子化に資するかについて説明した．