| GO CLARK KENDRICK CHENG | D, 中間発表 | 数理情報学 | 池田 和司, 笠原 正治, 久保 孝富, Nishanth Koganti |
|
Title: Analysing and Modelling Interactions Among Horses in a Herding Activity
Abstract: Many species across the animal kingdom exhibit different forms of collective motion under various circumstances. One specific example of collective motion among animals is herding in horses. In this talk, we focus on a family of Garrano horses, a breed of endangered feral horses in northern Portugal, and observe herding process of female horses by the stallion. We first qualitatively describe their herding behaviour while identifying the mechanisms along the way. Next, we infer and identify the different forces of interactions to be able to create a mathematical model of herding. We explain the parameters used and end with a discussion of their significance in creating this model.
Language of Presentation: English | |||
| ZHANG QI | M, 2回目発表 | 数理情報学 | 池田 和司, 笠原 正治, 吉本 潤一郎, 佐々木 博昭 |
|
title: Direct log-density gradient estimation with Gaussian mixture models and its application to clustering abstract: Some machine learning algorithms rely on the accurate estimation of probability density gradient, such as mode-seeking clustering, which assigns cluster labels by associating data samples with the nearest modes (zero points of the density gradient). We propose a method to estimate the gradient of the log-density which can be used for mode identification. Our work extends the least squares log-density gradient (LSLDG) by using Gaussian mixture models (GMMs), a method to directly estimate the gradient of the log-density without going through density estimation. The advantage of our work is that the correlation information in the gradient can be captured by GMMs, which leads to higher clustering accuracy. This method is then extended for hierarchical clustering. We show the validity via experiments. language of the presentation: English | |||
| QIU JIE | M, 2回目発表 | ロボティクス | 小笠原 司☆, 向川 康博, 金出 武雄(客員）, 高松 淳, 伍 洋 |
|
title: Skeleton Guided GANs for Person Re-identification
abstract: Person re-identification (ReID) plays a critical role in many video-surveillance applications. However, ReID faces many technical challenges which limit its performance and hinder its application in real systems. Pose variations caused by body movements and camera viewpoint changes are most significant among all these challenges. Though pose has already been concerned for ReID, all existing solutions have only used pose for alignment in feature extraction. In this paper, we explore and showcase a concrete example for a brand new direction: pose-adaptive whole sample generation for ReID. The model is named Skeleton Guided Deconvolutional-Generative Adversarial Network (SG-DGAN), which contains de-convolution networks to generate coarse images and generative adversarial networks to refine the image details for any given target pose. In order to preserve personal details in the target pose and make the generated images realistic, Siamese training framework and pre-definition loss are proposed in the refinement stage. For evaluation, we tested our framework on several public benchmark datasets to show the advanced performance of our framework. Experiments on representative large-scale benchmark datasets such as Market1501, Duke and CUHK03 demonstrate the superiority of our proposed SG-DGAN model, and a preliminary transfer learning experiment on VIPeR shows its encouraging generalization ability. language of the presentation: *** English *** | |||
| WATANAKEESUNTORN WASSAPON | M, 1回目発表 | ソフトウェア設計学 | 飯田 元 |
|
Title: OpenFlow Interactive Monitoring
Abstract: Software Defined Network (SDN) is a new approach of networking that emphasis on the network programmability and more dynamic control of the network. OpenFlow is the broadly used standard to implement the SDN network. However, understanding the dynamic behavior of an OpenFlow network is still challenging since the information about the operation is distributed across numerous network switches. This paper presents a tool call Opimon (OpenFlow Interactive Monitoring) that monitor and visualizes the OpenFlow network based SDN. Opimon provides a real-time monitoring and visualization of an OpenFlow network including network topology and flow tables of each switch. Web based user interface enables a user to quickly follows the network behavior and identify the problem. Opimon has been deployed successfully on a PRAGMA-ENT international testbed and enables researchers to understand more about the behavior of the testbed. Language of the presentation: English | |||
| KARIM MOHAMMAD BOZLUL | D, 中間発表 | 計算システムズ生物学 | 金谷 重彦, 松本 健一, MD.ALTAF-UL-AMIN, 小野 直亮 |
|
title:Clustering and analysis of Species-VOC Bipartite relations by a Novel Approach
Abstract: In the present work we propose a novel bi-clustering approach called BiClusO. Biclustering can be applied to various types of bipartite data e.g. gene-condition, gene-disease etc. We applied BiClusO to bipartite relations between species and Volatile organic compounds (VOCs). VOCs emitted by different species have huge environmental and ecological impacts. Biosynthesis of VOCs depends on different metabolic pathways based on which the species can be categorized. Previous study related to KNApSAcK VOC database was to classify microorganisms based on their VOC profiles which confirmed the consistency between VOC and pathogenicity based classification. However, classification of all species in terms of VOC profiles was not done due to limited data. We enriched our database with additional data collected from different online sources and journals. By applying BiClusO to species-VOC relational data we examined that VOC based classification is consistent with taxonomy based classification of the species. We further assessed the diversity of VOC pathways across different species classes. new the presentation: English * | |||
| KARIM MD. REJAUL | D, 中間発表 | ソフトウェア設計学 | 飯田 元, 松本 健一, 市川 昊平, 崔 恩瀞 |
|
title:Identifying and Predicting Key Features to Support Bug Reporting
abstract:Bug reports have primary means for developers to triage and fix bugs. In order to triage and fix the bugs effectively, bug reporters should clearly describe features in the bug reports that are important for the developers. Previous studies found that the reporters do not always provide the features that developers need for fixing the bugs. This study, first, performs an exploratory study to identify key features that the reporters frequently miss providing in their initial bug report submissions. Then, this study proposes an approach to predict whether the reporters should provide the key features to make a good bug report. As a case study using the bug reports for Camel, Derby, and Wicket projects, we found that Steps to Reproduce, Test Case, Code Example, Stack Traces, and Expected Behavior are the most additionally required features that the reporters frequently miss providing in their initial bug report submissions. We also found that the additionally required features significantly affect the bug-fixing process. Based on our findings, we build classification models using Naive Bayes (NB), Naive Bayes Multinomial (NBM), KNN (K-Nearest Neighbors), and Support Vector Machine (SVM) text classification techniques to predict key features by leveraging historical bug fixing knowledge. Our models achieve the best f1-score for Code Example, Test Case, Steps to Reproduce, Stack Trace, and Expected Behavior are 0.7 (Wicket), 0.7 (Derby), 0.57 (Derby), 0.57 (Derby), and 0.68 (Camel) respectively which are promising. Our contribution can benefit the reporters to improve the contents of the bug reports that are important for the developers to triage and fix the bugs effectively. language of the presentation: English | |||
| CHINTHANET BODIN | M, 2回目発表 | ソフトウェア工学 | 松本 健一, 飯田 元, 石尾 隆, Raula Gaikovina Kula |
|
title: How are known vulnerability fixes packaged and spread? A case study of npm JavaScript Ecosystem
abstract: In recent times, the vulnerability of library has become a big concern for the developer because of its impact on many packages in the ecosystem. The recent studies show that developers do not update the vulnerability. This behavior of developers may lead the ecosystem prone to risk to be attacked. I hypothesize that how vulnerability fixes packaged would affect the speed of releasing and spreading of vulnerability patch in the ecosystem. In this study, to test my hypothesis, I perform the empirical study to investigate how are known vulnerability fixes packaged. I first pick 10 vulnerabilities that refer to GitHub pull request and issue page and manually investigate how developers create the fix for the vulnerability and how they packaged them for the new release. I then analyze the target of vulnerability fixes from 188 fixed vulnerabilities. In this presentation, I would like to show the result of this study and suggestion for the developers in the ecosystem. For the future work, I would like to study more about how the fixes of vulnerability spread in the network of dependency and mitigate the risk to be attacked in the ecosystem. language of the presentation: English | |||