Companion Proceedings of the The Web Conference 2018最新文献_第9页

Training 100,000 Classes on a Single Titan X in 7 Hours or 15 Minutes with 25 Titan Xs 在7小时或15分钟内使用25个Titan X在单个Titan X上训练100,000个课程

Companion Proceedings of the The Web Conference 2018 Pub Date : 2018-04-23 DOI: 10.1145/3184558.3193135

Anshumali Shrivastava

{"title":"Training 100,000 Classes on a Single Titan X in 7 Hours or 15 Minutes with 25 Titan Xs","authors":"Anshumali Shrivastava","doi":"10.1145/3184558.3193135","DOIUrl":"https://doi.org/10.1145/3184558.3193135","url":null,"abstract":"In this talk, I will present Merged-Averaged Classifiers via Hashing (MACH) for K-classification with ultra-large values of K. Compared to traditional one-vs-all classifiers that require $O(Kd)$ memory and inference cost, MACH only need $O(dłogK)$ (d is dimensionality) memory while only requiring $O(KłogK + dłogK )$ operation for inference. MACH is a generic K-classification algorithm, with provably theoretical guarantees, without any assumption on the relationship between classes. MACH uses universal hashing to reduce classification with a large number of classes to few (logarithmic many) independent classification tasks with small (constant) number of classes. I will show the first quantification of discriminability-memory tradeoff in multi-class classification. Using the simple idea of hashing, we can train ODP dataset with 100,000 classes and 400,000 features on a single Titan X GPU, with the classification accuracy of 19.28%, which is the best-reported accuracy on this dataset. Before this work, the best performing baseline is a one-vs-all classifier that requires 40 billion parameters (160 GB model size) and achieves 9% accuracy. In contrast, MACH can achieve 9% accuracy with 480x reduction in the model size (of mere 0.3GB). With MACH, we also demonstrate complete training of feature extracted fine-grained imagenet dataset (compressed size 104GB), with 21,000 classes, on a single GPU. To the best of our knowledge, this is the first work to demonstrate complete training of these extreme-class datasets on a single Titan X. Furthermore, the algorithm is trivially parallelizable. Our experiments show that we can train ODP datasets in 7 hours on a single GPU or in 15 minutes with 25 GPUs. Similarly, we can train classifiers over the fine-grained imagenet dataset in 24 hours on a single GPU which can be reduced to little over 1 hour with 20 GPUs.","PeriodicalId":235572,"journal":{"name":"Companion Proceedings of the The Web Conference 2018","volume":"3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-04-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124358347","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Recoin: Relative Completeness in Wikidata Recoin:维基数据的相对完整性

Companion Proceedings of the The Web Conference 2018 Pub Date : 2018-04-23 DOI: 10.1145/3184558.3191641

Vevake Balaraman, Simon Razniewski, Werner Nutt

{"title":"Recoin: Relative Completeness in Wikidata","authors":"Vevake Balaraman, Simon Razniewski, Werner Nutt","doi":"10.1145/3184558.3191641","DOIUrl":"https://doi.org/10.1145/3184558.3191641","url":null,"abstract":"The collaborative knowledge base Wikidata is the central storage of Wikimedia projects, containing over 45 million data items. It acts as the hub for interlinking Wikipedia pages about a specific item in different languages, automates features such as infoboxes in Wikipedia, and is increasingly used for other applications such as data enrichment and question answering. Tracking the quality of Wikidata is an important issue for this project. In this paper we focus particularly on the completeness aspect. Several automated techniques have been adopted by Wikis to track and manage completeness, yet these techniques are generally subjective and do not provide a clear quality estimate at the level of entities. In this paper, we present an approach towards measuring Relative Completeness in Wikidata by comparison with data present for similar entities. This relative completeness approach is easily scalable with the introduction of new classes in the knowledge base, and has been implemented for all available entities in Wikidata. The results provide an intuition on the completeness of an entity comparing it with other similar entities. Here, we present our implementation approach along with a discussion on strategies and open challenges.","PeriodicalId":235572,"journal":{"name":"Companion Proceedings of the The Web Conference 2018","volume":"2017 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-04-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127547810","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 57

When Trust Saves Energy: A Reference Framework for Proof of Trust (PoT) Blockchains 当信任节省能源:信任证明(PoT)区块链的参考框架

Companion Proceedings of the The Web Conference 2018 Pub Date : 2018-04-23 DOI: 10.1145/3184558.3191553

Leila Bahri, Sarunas Girdzijauskas

引用次数: 29

THOR: Text-enabled Analytics for Humanitarian Operations THOR:人道主义行动文本分析

Companion Proceedings of the The Web Conference 2018 Pub Date : 2018-04-23 DOI: 10.1145/3184558.3186965

M. Kejriwal, Daniel Gilley, Pedro A. Szekely, J. Crisman

引用次数: 4

May I Have Your Attention, Please: - Building a Dystopian Attention Economy 请注意:-建立一个反乌托邦的注意力经济

Companion Proceedings of the The Web Conference 2018 Pub Date : 2018-04-23 DOI: 10.1145/3184558.3191605

S. Helmer

引用次数: 4

3rd EKM Workshop Chairs' Welcome & Organization 第三届EKM工作坊主席欢迎及组织

Companion Proceedings of the The Web Conference 2018 Pub Date : 2018-04-23 DOI: 10.1145/3184558.3192299

I. Lahoud, Elsa Cardoso, N. Matta

{"title":"3rd EKM Workshop Chairs' Welcome & Organization","authors":"I. Lahoud, Elsa Cardoso, N. Matta","doi":"10.1145/3184558.3192299","DOIUrl":"https://doi.org/10.1145/3184558.3192299","url":null,"abstract":"Welcome to the 3rd Educational Knowledge Management (EKM) workshop, which takes place at the WWW18 conference, in Lyon, France. The first edition was in 2014 in conjunction with the International Conference on Knowledge Engineering and Knowledge Management (EKAW), which was held in Linköping, Sweden, and the second one with EKAW 2016, in Bologna, Italy. We received 6 papers from all around the world covering a broad range of topics. Each paper was reviewed by three members of the program committee. After the reviewing process, three papers were accepted for inclusion into the WWW proceedings volume, two of them as full papers and one as a short paper. The EKM2018 workshop will run as a half-day event, including the papers' presentation and two invited speaker sessions. A Best Paper Award will be assigned, and authors will be invited to submit an extended version of their work to a special issue that will be published as part of the \"International Journal of Continuing Engineering Education and Lifelong Learning.\" The first paper Construction and Applications of TeKnowbase A Knowledge Base of Computer Science concepts' by Rajna Upadhyay, Ashutosh Bindal, Manjeet Kumar, and Maya Ramanath describes the development and evaluation of TeKnowbase, and how to use it in a variety of applications for learning a new topic, classification of technical text and querying and ranking computer science articles. The second one untitled \"Ontology-based recommender system in higher education\" by Charbel Obeid, Inaya Lahoud, Hicham El Khoury, and Pierre-Antoine Champin, is a position paper discussing an ontology-based recommender system to support a student's choice of major and university. The third paper \"Automatic Generation of Quizzes from DBpedia According to Educational Standards\" by Oscar Rodríguez Rocha and Catherine Faron Zucker focuses on educational quizzes. The authors present an approach to generate quizzes automatically from existing knowledge bases available on the Web of Linked Open Data (LOD), according to the official French educational standards.","PeriodicalId":235572,"journal":{"name":"Companion Proceedings of the The Web Conference 2018","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-04-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130399293","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

WWW'18 Open Challenge: Financial Opinion Mining and Question Answering WWW'18开放挑战:金融意见挖掘与问答

Companion Proceedings of the The Web Conference 2018 Pub Date : 2018-04-23 DOI: 10.1145/3184558.3192301

Macedo Maia, S. Handschuh, A. Freitas, Brian Davis, Ross McDermott, Manel Zarrouk, A. Balahur

引用次数: 115

Residual Convolutional LSTM for Tweet Count Prediction 残差卷积LSTM用于Tweet计数预测

Companion Proceedings of the The Web Conference 2018 Pub Date : 2018-04-23 DOI: 10.1145/3184558.3191571

Hong Wei, Hao Zhou, Jagan Sankaranarayanan, Sudipta Sengupta, H. Samet

引用次数: 18

Challenges Track Chairs' Welcome 挑战轨道椅欢迎

Companion Proceedings of the The Web Conference 2018 Pub Date : 2018-04-23 DOI: 10.1145/3184558.3192294

M. d’Aquin, Elena Cabrio

引用次数: 0

OSNED 2018 Chairs' Welcome & Organization OSNED 2018主席欢迎及组织

Companion Proceedings of the The Web Conference 2018 Pub Date : 2018-04-23 DOI: 10.1145/3184558.3192312

Rémy Cazabet, A. Passarella, Giulio Rossetti, F. Silvestri

引用次数: 0