| LIU HUAKUN | M, 2回目発表 | サイバネティクス・リアリティ工学 | 清川 清, 安本 慶一, 内山 英昭, Perusquia Hernandez Monica, 平尾 悠太郎 |
|
title: Improving Inertial-based Odometry via Deep IMU Online Calibration
abstract: This work presents a deep data-driven inertial measurement unit (IMU) online calibration method that can compensate for the run-time errors of the accelerometer and gyroscope to improve inertial-based odometry. We design a differential error learning strategy based on the kinematic motion model to train the sensor error compensation model. This strategy allows our method to learn IMU sensor errors, such as scale factors, axis-misalignment, and biases, solely from displacement and orientation increments given by external tracking systems. Then during the odometry computation, the trained model leverages the past inertial data to mitigate the sensor errors and thus reduces the integration errors to reflect the odometry state. The experiments conducted on two public visual-inertial datasets show an average of 20% improvement in the position estimation accuracy of visual-inertial odometry, which is comparable to existing learning-based methods with lower operational complexity. language of the presentation: English | |||
| ROCHA CONFESSOR BRIAN | M, 1回目発表 | サイバネティクス・リアリティ工学 | 清川 清, 荒牧 英治, 内山 英昭, Perusquia Hernandez Monica, 平尾 悠太郎 |
|
title: Therapeutic Applications of Location-Based Games on Hikikomori resocialization
abstract: The issue of extreme social isolation in Japan, known as hikikomori, has concerned the Japanese government for decades, and approaches to help affected individuals return to a normal social life have since been sought. In light of this, the current research aims to investigate the efficacy of Location-Based Games as a possible therapeutic tool to help said condition. For that, the development of an Location-Based mobile game with social elements in its design is proposed. To develop this game with a user-centric design, a survey is being done to investigate the game preferences of hikikomori. The results of this survey will be used to develop a mobile game that hikikomori participants will play in an experiment, and their reactions to the game will be measured to assess whether the intervention has any effect on their social isolation. language of the presentation: English | |||
| KESSY SUZAN JOSEPH | D, 中間発表 | 光メディアインタフェース | 向川 康博, 加藤 博一, 舩冨 卓哉, 藤村 友貴, 北野 和哉 |
|
title: *** Compensation of Temporal Illumination Variation on Whisk-broom Hyper-spectral Imaging *** abstract: *** This work presents a novel method for compensating temporal variations in whisk-broom hyperspectral imaging. Whiskbroom imaging scans the scene sequentially, capturing a complete spectrum at each spatial coordinate point-by-point over time. While lengthy measurement times are not problematic without temporal light fluctuations, capturing outdoor cultural artefacts often involves time-varying illumination, resulting in varying measured values for the same scene at different times. Previous approaches incorporated an extra single-vertical scan alongside the standard raster (horizontal) scan for compensation. However, errors increased when the extra scan was performed near or on a black frame. To overcome this limitation, we propose a method that incorporates multiple columns or a full-vertical scan in addition to the horizontal scan. Furthermore, we introduce logarithm space to the formulation and utilise the low-dimensional models of logarithms of spectra for this problem. This logarithm-based approach transforms multiplications into additions, enabling the creation of linear equations for efficient and analytical solutions. The proposed method effectively eliminates the variations in illumination. To highlight the practical utility of our method, we successfully apply it to capture hyperspectral images of the historic stained-glass windows in the Amiens Cathedral, France. *** language of the presentation: *** English *** | |||
| ZHANG FAN | D, 中間発表 | 計算システムズ生物学 | 金谷 重彦, 宮尾 知幸, 黄 銘, MD.ALTAF-UL-AMIN, 小野 直亮 |
|
title: Interpretation of Gaussian Process using integrated gradients abstract: In chemoinformatics, in order to efficiently find compounds with desired properties and activities from the vast chemical space, Gaussian processes which can calculate not only predictions but also prediction reliability is often used. Although the Gaussian process is difficult to interpret because it is a nonparametric machine learning model. In this study, we propose a method to interpret not only the predictions but also the prediction reliability of Gaussian processes using integrated gradients. Using the proposed method, we confirmed that not only the Gaussian process but also Bayesian optimization and deep kernel learning Gaussian processes can be interpreted. language of the presentation: Japanese | |||
| YANG SHUO | M, 2回目発表 | 計算システムズ生物学 | 金谷 重彦, 峠 隆之, 黄 銘, MD.ALTAF-UL-AMIN, 小野 直亮 |
|
title: Clustering-based Natural Products Antibacterial activity prediction
abstract: With the gradual increase in drug resistance of bacteria, viruses and fungi, the search for new antimicrobial agents has become an urgent need. Natural products are widely considered to have potential antimicrobial properties. In this study, cluster analysis was used to predict the antimicrobial properties of natural products. Through cluster analysis, we clustered natural product molecules based on their structural and chemical characteristics. The results showed structural and chemical similarities between certain natural products and known antimicrobial drugs, indicating that they may have antimicrobial activity and be potential drug candidates. language of the presentation: English | |||
| MUHAMMAD ALQAAF SUBANDOKO | M, 2回目発表 | 計算システムズ生物学 | 金谷 重彦, 峠 隆之, MD.ALTAF-UL-AMIN, 黄 銘, 小野 直亮 |
|
TITLE: Discovering Natural Anti-COVID Agents: A Combined Sequence and Docking Study
ABSTRACT: In this study, we pursued the discovery of natural compounds as potential therapeutic agents against COVID-19, leveraging a computational approach that combines sequence analysis and molecular docking. We analyzed 204 unique spike protein sequences from the SARS-CoV-2 virus and identified five distinct classes using a clustering algorithm. We utilized BindingDB to investigate protein-small molecule interactions, creating a comprehensive dataset of binding information across diverse species. Further, we compared the binding molecules with secondary metabolites from the KNApSAcK database, using the Tanimoto similarity method and sequence identity as selection criteria. A molecular docking analysis was performed to validate and rank potential metabolites based on their binding affinity to spike proteins. Our results yielded promising natural compounds that exhibit high binding affinity, fostering future in-depth investigations to validate these potential anti-COVID-19 agents. LANGUAGE: English (英語) | |||
| 井阪 友哉 | M, 2回目発表 | ディペンダブルシステム学 | 井上 美智子, 中島 康彦, 新谷 道広 |
|
title: Hyperdimensional Computing Platform for Low Power Devices
abstract: Hyperdimensional computing (HDC) can perform various cognitive tasks efficiently by mapping data to hyperdimensional vectors consisting of thousands to tens of thousands of dimensions. On the other hand, since the main operations of HDC, Bind, Permutation and Bound, require several cycles in the computing unit, it is not necessarily efficient to perform HDC on CPU platform. In this paper, we propose a computing platform specialized for HDC. Our platform can execute various tasks at high speed and with low power consumption by cooperating with the CPU. Furthermore, by making it possible to freely select three operations per cycle, our accelerator enables to support any HDC encoding method. Through evaluation experiments with the ARM-v7 processor, we show that the above operations can be accelerated by a maximum of 169 times. We also confirmed that our accelerator can improve the energy-delay product up to 13,469 times in the training of handwritten character recognition task. language of the presentation: Japanese | |||