News

The 31th Annual Conference of the Japanese Society for Artificial Intelligence, 2017 (JSAI 2017) (DOMESTIC)

Received Excellent Paper Award

May 26, 2017

The following papers won the Excellent Paper Award at JSAI 2017(external link ).

乗換案内データを用いた未来の混雑予測の研究 Kota Tsubouchi, Masamichi Shimosaka(Tokyo Institute of Technology), Tatsuya Konishi(The University of Tokyo), Mikiya Maruyama and Tatsuo Yamashita

The results of this research are already being exploited in Yahoo! JAPAN Transit. We intend to continue our research on from here, always aiming at a better user experience.

AdaFlock:予測モデリングのためのクラウドソーシングによる適応的特徴生成法 Ryusuke Takahama (scouty Inc.)*, Yukino Baba (Kyoto University), Nobuyuki Shimizu, Sumio Fujita and Hisashi Kashima (Kyoto University)

The paper describes research results from the joint research project with JST-ERATO Large Graph Project(external link). It proposes “AdaFlock”, algorithm to generate effective feature sets for various machine learning procedures by leveraging information from crowdsourced tasks. The following related paper also accepted as a full research paper to appear in AAAI 2018, one of the most prestigious conferences in the AI domain.

AdaFlock: Adaptive Feature Discovery for Human-in-the-loop Predictive Modeling Ryusuke Takahama (scouty Inc.)*, Yukino Baba (Kyoto University), Nobuyuki Shimizu, Sumio Fujita and Hisashi Kashima (Kyoto University)

The investigated methodologies will be adopted in many machine learning tasks of our diverse services, especially where feature designing by human experts is hard due to vast and complex natures of data.

*: This work was carried out while the first author worked at JST,ERATO, Kawarabayashi Large Graph Project.