A Research on Big Data and AI Analysis Algorithm Optimization Using GPUs

• Item type: Symposium(1)
• 公開日: 2018-06-27
• タイトル: A Research on Big Data and AI Analysis Algorithm Optimization Using GPUs
• 言語: eng
• キーワード / 主題Scheme: Other
• キーワード / 主題: その他
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_5794
• 資源タイプ: conference paper
• 著者所属: 東京大学情報理工学系研究科コンピュータ科学専攻中田研究室
• 著者所属: 東京大学情報理工学系研究科ソーシャルICT研究センター
• 著者所属: 東京大学情報理工学系研究科ソーシャルICT研究センター
• 著者所属: 東京大学情報理工学系研究科コンピュータ科学専攻兼ソーシャルICT研究センター
• 著者名: Van, Sang Tran; 小林, 良輔; 山口, 利恵; 中田, 登志之
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: With the significant increase of computer performance, in recent years, many complex human-like tasks have been resolved by computer software in reasonable time. These tasks include visual object recognition, speech to text interpretation, human face authentication, etc... However, computer performance is going to reach the limit as the CMOS transistor size near the limit. On the other hand, the amount of data which need to be processed is incredibly growing up under Internet of Thing, Industrialization 4.0, social network era, which leads to the demand of higher scalability on current Big Data, AI analysis algorithms. Our research investigated on finding scaling solution for Big Data, AI analysis problems. The whole development is composed of 2 phases: acceleration by GPU and distributed computing application. This research focuses on the former topic. A real-world dataset was used in this research to achieve more real-life optimization and model evaluation result.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: With the significant increase of computer performance, in recent years, many complex human-like tasks have been resolved by computer software in reasonable time. These tasks include visual object recognition, speech to text interpretation, human face authentication, etc... However, computer performance is going to reach the limit as the CMOS transistor size near the limit. On the other hand, the amount of data which need to be processed is incredibly growing up under Internet of Thing, Industrialization 4.0, social network era, which leads to the demand of higher scalability on current Big Data, AI analysis algorithms. Our research investigated on finding scaling solution for Big Data, AI analysis problems. The whole development is composed of 2 phases: acceleration by GPU and distributed computing application. This research focuses on the former topic. A real-world dataset was used in this research to achieve more real-life optimization and model evaluation result.
• 書誌情報: マルチメディア，分散協調とモバイルシンポジウム2018論文集; 巻 2018, p. 1212-1219, 発行日 2018-06-27
• 出版者 / 言語: ja
• 出版者: 情報処理学会