single-award.php
JACIII Best Paper and Young Researcher Awards 2024
 |
JACIII BEST PAPER AWARD 2024 Toward Question-Answering with Multi-Hop Reasoning and Calculation over Knowledge Using a Neural Network Model with External Memories Yuri Murayama and Ichiro Kobayashi |
 |  |  |
| Yuri Murayama Ochanomizu University | Ichiro Kobayashi Ochanomizu University |
Photo: From Prof. Kaoru Hirota, the JACIII Editor-in-Chief, Prof. Ichiro Kobayashi and Dr. Yuri Murayama received a plaque and certificates of the JACIII Best Paper Award 2024.
Message from the Winner
We are deeply honored to receive the JACIII Best Paper Award 2024. We would like to express our sincere gratitude to the reviewers, JACIII editorial office, and Fuji Technology Press for their generous support. The differentiable neural computer (DNC) proposed by Graves et al. is a neural network model with addressable external memory that can solve algorithmic and question-answering tasks. The DNC realizes question answering with premises by writing an input sequence into the memory and reading the necessary information to infer the answer from the memory. However, integrating structured knowledge and calculations into DNC models remains challenging. In our study, we incorporated an architecture for knowledge and calculations into DNC models to improve their ability to generate correct answers to questions with multi-hop reasoning and to provide calculations over structured knowledge. Consequently, our improved DNC models achieved the best accuracy for complex questions requiring multi-hop reasoning and proper calculations over knowledge. We hope that our work will contribute to the field of artificial intelligence research and that this area will continue to advance.
|
JACIII YOUNG RESEARCHER AWARD 2024 Human Pose Estimation with Multi-Camera Localization Using Multi-Objective Optimization Based on Topological Structured Learning Takenori Obo, Kunikazu Hamada, Masatoshi Eguchi, and Naoyuki Kubota |
 |  |  |  | |
| Takenori Obo Tokyo Metropolitan University | Kunikazu Hamada Tokyo Metropolitan University | Masatoshi Eguchi Tokyo Metropolitan University | Naoyuki Kubota Tokyo Metropolitan University |
Photo: Dr. Takenori Obo received a certificate and a plaque of the JACIII Young Researcher Award 2024.
Message from the Winner
I am deeply honored to receive the “JACIII Young Researcher Award 2024.” I sincerely thank the selection committee, reviewers, and co-researchers for their invaluable assistance.
The winning paper, “Human Pose Estimation with Multi-Camera Localization Using Multi-Objective Optimization Based on Topological Structured Learning,” proposes a method for 3D pose estimation that uses joint angle estimation and self-localization of each camera using multiple smart devices equipped with 2D skeleton detection libraries. This study uses a multi-island genetic algorithm to optimize the joint angles of a kinematic model representing human posture, as well as the relative angles of each camera from the subject’s perspective. In addition, to improve the genetic algorithm’s search efficiency, we implemented a structured learning method based on topological mapping. This method maintains the estimated poses as reference vectors, allowing the genetic algorithm to use them as search points.
Marker-less motion capture systems used in rehabilitation and caregiving settings are frequently expensive, making it challenging to analyze human motion easily. We believe that the proposed system will provide significant assistance in these areas. This award motivated me to continue working in practice, research, and education, with the goal of contributing to society, developing talent, and advancing academic fields. I would like to express my heartfelt gratitude once more for this award.