===============================================
Technical Committee on Infomation-Based Induction Sciences and Machine Learning (IBISML)
Chair: Hisashi Kashima (Kyoto Univ.)
Vice Chair: Masashi Sugiyama (Univ. of Tokyo), Koji Tsuda (Univ. of Tokyo)
Secretary: Ichiro Takeuchi (Nagoya Inst. of Tech.), Toshihiro Kamishima (AIST)
Assistant: Tomoharu Iwata (NTT), Shigeyuki Oba (Kyoto Univ.)
DATE:
Tue, Mar 5, 2019 13:30 - 17:30
Wed, Mar 6, 2019 10:00 - 14:30
PLACE:
RIKEN Center for Advanced Intelligence Project(1-4-1 Nihombashi Chuo-ku, 103-0027 Tokyo, Japan. Direct access from Tokyo Metro Nihombashi Station. https://aip.riken.jp/access/?lang=en. Koji Tsuda. +81-3-6225-2373)
TOPICS:
Machine learning, etc.
----------------------------------------
Tue, Mar 5 PM Theory of Machine Learning (13:30 - 16:30)
Chair: Koji Tsuda
----------------------------------------
(1) 13:30 - 14:00
A connection between Kida's optimal signal approximation and the signal classification by maximizing margin
Takuro Kida (Tokyo Tech. Prof.EM), Yuichi Kida (Ohu Univ.)
(2) 14:00 - 14:30
Expressive power of skip connection and network architecture
Jumpei Nagase, Tetsuya Ishiwata (Shibaura Inst. of Tech.)
(3) 14:30 - 15:00
Efficient Exploration by Variational Information Maximizing Exploration on Reinforcement Learning
Kazuki Doi, Keigo Okawa (Gifu Univ.), Motoki Shiga (Gifu Univ./JST/RIKEN)
----- Break ( 15 min. ) -----
(4) 15:15 - 16:15
Invited Talk
Hiroshi Maruyama (Preferred Networks)
Invitation to Machine Learning Engineering
----- Break ( 15 min. ) -----
----------------------------------------
Tue, Mar 5 PM Application of Machine Learning (16:30 - 17:30)
Chair: Koji Tsuda
----------------------------------------
(5) 16:30 - 17:00
Set transformer for coordinating outfits
Takuma Nakamura, Yuki Saito (ZOZO Research)
(6) 17:00 - 17:30
A Study on Automatic Generation of False Data Injection Attack against Connected Car Service Based on Reinforcement Learning
Yuichiro Dan, Keita Hasegawa, Takafumi Harada, Tomoaki Washio, Yoshihito Oshima (NTT)
----------------------------------------
Wed, Mar 6 AM Theory of Machine Learning (10:00 - 13:00)
Chair: Koji Tsuda
----------------------------------------
(7) 10:00 - 10:30
Effects of Batch-normalization on Fisher Information Matrix of ResNet
Yasutaka Furusho, Kazushi Ikeda (NAIST)
(8) 10:30 - 11:00
Wider neural networks with ReLU activation generalize better
Yasutaka Furusho, Kazushi Ikeda (NAIST)
(9) 11:00 - 11:30
Shapelet-based Multiple-Instance Learning
Daiki Suehiro, Kohei Hatano (Kyushu Univ./RIKEN AIP), Eiji Takimoto (Kyushu Univ.), Shuji Yamamoto, Kenichi Bannai (Keio Univ./RIKEN AIP), Akiko Takeda (The Univ. of Tokyo/RIKEN AIP)
(10) 11:30 - 12:00
Optimal Kernel for Mode Estimation via Kernel Density Estimation
Ryoya Yamasaki, Toshiyuki Tanaka (Kyoto Univ.)
----- Lunch ( 60 min. ) -----
----------------------------------------
Wed, Mar 6 PM Theory and Application of Machine Learning (13:00 - 14:30)
Chair: Toshihiro Kamishima
----------------------------------------
(11) 13:00 - 13:30
Magnetic Resonance Angiography Image Restoration by Super Resolution based on Deep Learning
Shizen Kitazaki, Masanori Kawakita, Yutaka Jitumatu (Kyushu Univ.), Shigehide Kuhara (Kyorin Univ.), Akio Hiwatashi, Jun'ichi Takeuchi (Kyushu Univ.)
(12) 13:30 - 14:00
Acceleration of Boosting Discriminators Using Region Partition Learning and Its Application to Face Detectors
Takeshi Mori, Junichi Takeuchi, Masanori Kawakita (Kyushu Univ.)
(13) 14:00 - 14:30
Evaluation of LGRF using RAVE in Go program
Tatsuya Shimizu, Yuta Hayama, Asuka Nakamura, Yoshitaka Maekawa (CIT)
# Information for speakers
General Talk will have 25 minutes for presentation and 5 minutes for discussion.
=== Technical Committee on Infomation-Based Induction Sciences and Machine Learning (IBISML) ===
# FUTURE SCHEDULE:
Mon, Jun 17, 2019 - Wed, Jun 19, 2019: Okinawa Institute of Science and Technology [Thu, May 2], Topics: Neurocomputing, Machine Learning Approach to Biodata Mining, and General
Last modified: 2019-03-05 15:27:18
|
Notification: Mail addresses are partially hidden against SPAM.
|