ゼミナール発表

日時: 7月19日(水)3限 (13:30-15:00)


会場: L1

司会: 諏訪 博彦
ANTONIO VICTOR ANDREW ASUNCION 1561204: D, 中間発表 計算システムズ生物学 金谷 重彦, 佐藤 嘉伸, MD.ALTAF-UL-AMIN, 小野 直亮
Title: A Convolutional Neural Network Approach to the Classification and Reconstruction of Biomedical Images Abstract: Lung cancer can be fatal if it is not treated immediately. Recently, there has been a lot of development in computer-aided diagnosis that focuses on lung cancer. Our goal is to use convolutional neural architecture as the foundation for several networks whose purpose will be the reconstruction and classification of lung cancer images, with the purpose of understanding the mechanics of both the images and the networks themselves. A dataset from The Cancer Genome Atlas database will be used as input for two types of autoencoder networks and a standard classifier network. We also apply experiments regarding the convolutional filter size. We show that for optimal results, we choose a smaller window for reconstruction, and a larger window for classification. We also look into some visualization examples to understand the inner mechanics of the networks. Language: English
 
尾上 紗野 1561008: D, 中間発表 ソフトウェア工学 松本 健一, 岡田 実, 安本 慶一, 門田 暁人, 畑 秀明
title: Characteristics of Participants Activities and Development Communities in OSS Projects
abstract: Background: Open Source Software (OSS) projects are important part of Software Development. Although some of the hosted projects are growing and have many developers, most projects are organized by a few developers and face difficulties in terms of sustainability. OSS projects depend mainly on volunteer contributors, and attracting and retaining these volunteers are major concerns of the project stakeholders. Aims: This study understands the state of OSS projects from different perspectives of health. We present two studies of understand these perspectives. In the field of demography or economics, the health is considered to understand the state of a country. Also, understanding the state of OSS projects is essential to analyze features and predict the feature of projects. Method: We first present investigating the population structures of OSS development communities in detail and conduct software analytics to obtain actionable information using my proposed software population pyramid which is consist of coding contributors and non-coding contributors. We apply a demographic approach and predict population of an OSS project with the well-known cohort component method. We also present our current progress that is investigating state of OSS projects using the field of demography and economics. We apply this field to OSS projects to understand the communities and the evolution of OSS projects from the perspectives of health and wealth. We define the health is the community activities such as a contributor work rate, and the wealth is the product evolution. Results: First, we observe that: (1) there are four types of population structures in OSS development communities in terms of experiences and contributions; and (2) a cohort component population projection method can predict the future population accurately. In the second study, we observe that: (1) wealthy projects attract and rely on the casual workforce; and (2) less wealthy projects may require additional efforts from their more experienced contributors. Conclusions: In this study, we apply demographic and economic perspectives to OSS development community. Based on our research, we have proposed software population pyramid.
language of the presentation: Japanese
 

会場: L2

司会: 能地 宏
NGUYEN VAN KHANH 1651205: M, 1回目発表 知能コミュニケーション 中村 哲, 安本 慶一, 鈴木 優
Title: A weather-aware point-of-interest recommender system
Abstract:
A Tourism recommender systems (TRS) aims to give a list of Point-of-interests (POIs) that should be proper for the user in a particular context. In recent years, there is not much attention on discovering the impact of weather for TRS. In this paper, we contribute to this research field by presenting a method that applies the POI-weather-rating data to make a list of POI suitable for a certain type of weather. Our system ranks the POI according to a score function, where the score of each POI will depend on the average rating of users to different weather aspects, for example, precipitation probability, temperature. The results show that recommended POIs satisfy user with many weather condition, and the recommendation accuracy is better in comparison to the original algorithm.
Language of the presentation: English
 
JOHANES EFFENDI THE 1651208: M, 1回目発表 知能コミュニケーション 中村 哲, 松本 裕治, 須藤 克仁, Sakriani Sakit
Title: Multi Operation Paraphrase Generation

Abstract:
Paraphrasing plays an important role in many language generation task. Some technologies in language generation such as machine translation, abstractive summarization, and question answering need paraphrase to enrich and adds flexibility. The existing paraphrase corpus usually only have one-to-one parallel sentences. Furthermore, it's difficult to track the operation applied to the original sentence. In this research, we collect multi paraphrase corpus using crowdsourcing based on multi paraphrase operation which enables the possibility to choose which operation we need depending on the task. Based on this corpus, we also successfully developed multi-operation sequence-to-sequence model for automatic paraphrase generation. In this seminar, I will present about the motivation, purpose, and method about this research, and preliminary experiment on our sequence-sequence paraphrase model with the dataset. I will also explain about the future task of this research.

Language of the presentation: English
 
NGUYEN THE TUNG 1551205: M, 2回目発表 知能コミュニケーション 中村 哲, 松本 裕治, 吉野 幸一郎
title: Considering deception information in negotiation dialog
abstract: Negotiation has received a lot of attention from dialog research community recently. Existing negotiation systems give responses to user with the assumption that all user's arguments are honest. However, in real life negotiation, there are situations where the negotiators can use deceptive tactics to achieve the outcome that is more beneficial to them. To counter this problem, I proposed a dialog system that can detect user's lies by using multi-modal approach. In addition, a strategy of how to react when user is lying is also created. The deception information is considered as part of the dialog belief state, as a result; the system's dialog manager will choose the best response by considering both information about user's deception and user's dialog action. The policy of the dialog manager was trained using combination of POMDP and rule-based method.
language of the presentation: English
 
磯 颯 1651011: M, 2回目発表 自然言語処理学 松本 裕治, 荒牧 英治, 中村 哲, 進藤 裕之

title: Forecasting Word Model: Twitter-based Influenza Surveillance and Prediction

abstract: Because of the increasing popularity of social media, much information has been shared on the internet, enabling social media users to understand various real world events.

Particularly, social media-based infectious disease surveillance has attracted increasing attention.

In this work, we specifically examine influenza: a common topic of communication on social media.

The fundamental theory of this work is that several words, such as symptom words (\textit{fever}, \textit{headache}, etc.), appear in advance of flu epidemic occurrence.

Consequently, past word occurrence can contribute to estimation of the number of current patients.

To employ such forecasting words, one can first estimate the optimal time lag for each word based on their cross correlation.

Then one can build a linear model consisting of word frequencies at different time points for nowcasting and for forecasting influenza epidemics.

Experimentally obtained results (using 7.7 million tweets of August 2012 -- January 2016), the proposed model achieved the best nowcasting performance to date (correlation ratio $0.93$) and practically sufficient forecasting performance (correlation ratio $0.91$ in 1-week future prediction, and correlation ratio $0.77$ in 3-weeks future prediction).

This report reveals the effectiveness of the word time shift to predict of future epidemics using Twitter.


language of the presentation: Japanese

 

会場: L3

司会: 川上 朋也
PIPATANAKUL KUNAT 1551206: M, 2回目発表 サイバネティクス・リアリティ工学 清川 清, 加藤 博一, 佐藤 智和, 河合 紀彦
title: Indirect Augmented Reality with Online Panoramic Image Generation
abstract: Indirect Augmented Reality (IAR) is an approach which enables jitterless high-quality Augmented Reality experience with only rough estimation of device poses by using an omnidirectional image and superimposing virtual objects onto it in advance. However, existing IAR methods have a problem that they cannot be used in locations where omnidirectional images are not pre-captured. In this study, we introduce a new IAR method which simultaneously creates a panoramic image by stitching images captured by a monocular camera and superimposes virtual objects to realize jitterless AR experience on demand.
language of the presentation: English
 
RONGSIRIGUL THIWAT 1551207: M, 2回目発表 サイバネティクス・リアリティ工学 清川 清, 加藤 博一, 佐藤 智和, 中島 悠太
title: Novel View Synthesis with Light-weight View-dependent Texture Mapping for a Stereoscopic HMD
abstract: The proliferation of off-the-shelf head-mounted displays (HMDs) let end-users enjoy virtual reality applications, some of which render a real-world scene using a novel view synthesis (NVS) technique. View-dependent texture mapping (VDTM) has been studied for NVS due to its photo-realistic quality. The VDTM technique renders a novel view by adaptively selecting textures from the most appropriate images. However, this process is computationally expensive because VDTM scans every captured image. For stereoscopic HMDs, the situation is much worse because we need to render novel views once for each eye, almost doubling the cost. This paper proposes light-weight VDTM tailored for an HMD. In order to reduce the computational cost in VDTM, our method leverages the overlapping fields of view between a stereoscopic pair of HMD images and pruning the images to be scanned. We show that the proposed method drastically accelerates the VDTM process without spoiling the image quality through a user study.
language of the presentation: English
 
ZHANG QI 1651211: M, 1回目発表 数理情報学 池田 和司
title: Stable Kernel Density Estimation
abstract: Kernel Density Estimation (KDE) is a commonly used non-parametric method for density estimation task, which involves minimizing an empirical risk as an estimate of the real risk which is not available in most cases. However, in the case of sample size is small, the empirical risk shows large variance, so that the minima of it is useless. We propose to use a robust estimate of risk derived by m-estimator, then build the density estimation model by minimizing this robust risk. Simulation result shows that our proposed method has smaller standard deviation than KDE, real data test is left as our future work.
language of the presentation: English