Latifur Khan, Department of Computer Science, University of Texas at Dallas (UT Dallas), USA
Date: 23 July 2020
Time: 8:00 PM
Venue: Online, over Zoom
Title: MultiCon: A Multi-Contrastive Learning based Semi-Supervised Classification Framework and Its Applications Towards Covid19
Abstract: Deep neural networks (DNN) require a large number of an-notations, which sometimes is very expensive and cumber-some. Over the years, various efforts have been proposed forreducing the annotation cost when training the DNN. Semi-Supervised Learning (SSL) is one of the solutions that has beenprovably handy in leveraging unlabeled instances to mitigatethe efficacy of the model’s performance and has been attract-ing an increasing amount of attention in recent times. In thiswork, our main insight is that semi-supervised learning canbenefit from recently proposed unsupervised contrastive learn-ing approach, which aims to achieve the positive concentratedand negative separated representation in the unlabeled fea-ture space. Herein, we introduce MultiCon, a semi-supervisedlearning paradigm that aims at learning data augmentation in-variant based embedding. In particular, we combine the multi-contrastive learning approach with a consistency regularizationmethod for maximizing the similarity between differently aug-mented views of one sample and pushing the embedding of dif-ferent instances away in the latent space simultaneously. Exper-iments on multiple standard datasets including Covid19 ChestX-ray images and CT Scans demonstrate that MultiCon achievesstate-of-the-art performance across existing SSL benchmarks.This work is funded by NSF and NIH.
Bio: Dr. Latifur Khan is currently a full Professor (tenured) in the Computer Science depart-ment at the University of Texas at Dallas, USA where he has been teaching and conductingresearch since September 2000. He received his Ph.D. degree in Computer Science fromthe University of Southern California (USC) in August of 2000. Dr. Khan obtained hisB.Sc. degree in Computer Science and Engineering from Bangladesh University of Engi-neering and Technology (BUET), Dhaka, Bangladesh in November, 1993 with First classHonors (2nd position). He was a recipient of Chancellor Awards from the President ofBangladesh.Dr. Khan is an ACM Distinguished Scientist and received IEEE Big Data Security SeniorResearch Award, in May 2019, and Fellow of SIRI (Society of Information Reuse andIntegration) award in Aug, 2018. He has received prestigious awards including the IEEETechnical Achievement Award for Intelligence and Security Informatics, IEEE Big Data Security Award and IBMFaculty Award (research) 2016. Dr. Latifur Khan has published over 300 papers in premier journals such as VLDB,Journal of Web Semantics, IEEE TDKE, IEEE TDSC, IEEE TSMC, and AI Research and in prestigious conferencessuch as AAAI, IJCAI, ACM WWW, CIKM, ICDE, ACM GIS, IEEE ICDM, IEEE BigData, ECML/PKDD, PAKDD, ACMMultimedia, ICWC, ACM SACMAT, IEEE ICSC, IEEE Cloud and INFOCOM. He has been invited to give keynotesand invited talks at a number of conferences hosted by IEEE and ACM. In addition, he has conducted tutorialsessions in prominent conferences such as SIGKDD 2017, 2016, IJCAI 2017, AAAI 2017, SDM 2017, PAKDD 2011& 2012, DASFAA 2012, ACM WWW 2005, MIS2005, and DASFAA 2007.Currently, Dr. Khan’s research area focuses on big data management and analytics, data mining and its applicationover cyber security, complex data management including geo-spatial data and multimedia data. His research hasbeen supported by grants from NSF, NIH, the Air Force Office of Scientific Research (AFOSR), DOE, NSA, IBM andHPE. More details can be found at:www.utdallas.edu/~lkhan/