Schedule
All times are in IST
June 18
June 18
2.15 p.m.–2.25 p.m.
Welcome Address
Welcome Address
3.00 p.m.–3.30 p.m.
Abstract:Abstract: This talk is motivated by the following thesis: Background knowledge is key to intelligent decision making. While deep learning methods have made significant strides over the last few years, they often lack the context in which they operate. Knowledge Graphs (and more generally multi-relational graphs) provide a flexible framework to capture and represent knowledge of various kinds, viz., factual, syntactic, and temporal. In this talk, I shall present an overview of my research on how to build and embed Knowledge Graphs, and how they can be effectively utilized within deep learning systems.
3:30 p.m.– 3:45 p.m.
BREAK
3.45 p.m.– 4:15 p.m.
Abstract: Hypothesis testing is a key problem in statistics with applications in almost all branches of scientific research. Hypothesis testing has recently become more prominent in AI and machine learning due to its connection to classification and clustering problems. In this talk, we briefly introduce the problem of hypothesis testing, particularly two-sample testing. We then focus on a class of hypothesis testing problems – testing of large sparse networks. This problem frequently arises in bioinformatics and computational neuroscience. We discuss the main challenge in solving this problem, and then develop a theory for testing a small collection of large graphs. Our theoretical analysis shows that it is possible to test large inhomogeneous Erdös-Rényi graph models with access to few random graphs, and the corresponding methods can be used in practice. This work was done in collaboration with Maurilio Gutzeit, Alexandra Carpentier, and Ulrike von Luxburg, and has been published in Neurips 2018 and The Annals of Statistics. Bio: Debarghya Ghoshdastidar conducts research in the theory of machine learning, artificial intelligence and network science. The main focus of his research is on the statistical understanding and interpretability of methods used in machine learning. His works provide new insights and algorithms for decision problems, involving complex data such as networks and preference relations, that arise in various fields including neuroscience, crowdsourcing and computer vision. After completing his PhD in 2016 at the Indian Institute of Science, he joined the University of Tuebingen as a post-doctoral researcher and led a junior research group funded by the Baden-Wuerttemberg Foundation. In 2019, he joined the Department of Informatics as an Assistant Professor for Theoretical Foundations of Artificial Intelligence.
4.15 p.m.–4.45 p.m.
Abstract:Reinforcement learning is a machine learning paradigm where the learning agent learns via direct interaction with the environment. The behaviour of the learning agent is captured in the so-called value function. The temporal difference (TD) class of algorithms are used for value function learning. Stability and rate of convergence of these algorithms depend on the choice of the step-size which is usually tuned in a problem specific manner. We show that it is possible to eliminate step-size tuning by choosing a constant step-size that is independent of the problem and with averaging of iterates, an optimal rate of O(1/t) for the convergence of mean-squared estimation error can be achieved.
Bio:Chandrashekar Lakshminarayanan’s areas are reinforcement learning, stochastic control and deep learning. He obtained his PhD from the Department of Computer Science and Automation, Indian Institute of Science (2016), and was a post-doctoral research fellow at the Department of Computing Science (July 2016- June 2017), University of Alberta, and a research scientist at DeepMind, London (August 2017-July 2018). Prior to his PhD, he was an analog design engineer at Cosmic Circuits, Bangalore for a period of 3 years. He joined IITPKD as an assistant professor in July 2018.
Bio:Chandrashekar Lakshminarayanan’s areas are reinforcement learning, stochastic control and deep learning. He obtained his PhD from the Department of Computer Science and Automation, Indian Institute of Science (2016), and was a post-doctoral research fellow at the Department of Computing Science (July 2016- June 2017), University of Alberta, and a research scientist at DeepMind, London (August 2017-July 2018). Prior to his PhD, he was an analog design engineer at Cosmic Circuits, Bangalore for a period of 3 years. He joined IITPKD as an assistant professor in July 2018.
4.45 p.m.–5.15 p.m.
Abstract:Learning discrete-valued functions from training data is the most popular machine learning task with practitioners. It includes the standard tasks of binary classification, multiclass classification, structured prediction, and ranking. In most practical scenarios, the ultimate task falls into one of these bins, but the evaluation metric of choice has to be specialized in some way, and it is often done via cost-sensitive classification, which is reasonably well studied. However, in several such scenarios, any "simple" or "decomposable" or "linear" evaluation metric is unacceptable due to situations such as unbalanced data and trivial solutions, paving the way for "complex" performance measures like the F-measure, or Harmonic-mean measure. Another face of complexity is in situations where the learnt classifier has to satisfy certain constraints, e.g. it must obey fairness by accepting an equal proportion of men and women to a college. We call such tasks with "complex" objectives as complex classification problems, and we will discuss a broad class of algorithms that have emerged to tackle problems of this kind.
Bio:Dr. Harish is currently an assistant professor at the computer science and engineering (CSE) department of IIT Madras. His primary areas of interest are in machine learning, statistical learning theory and optimisation. He was previously a research scientist at IBM research labs and a post-doc at the University of Michigan. He completed his Ph.D. at the Computer Science and Automation (CSA) department of the Indian Institute of Science (IISc), Bangalore advised by Prof. Shivani Agarwal.
Bio:Dr. Harish is currently an assistant professor at the computer science and engineering (CSE) department of IIT Madras. His primary areas of interest are in machine learning, statistical learning theory and optimisation. He was previously a research scientist at IBM research labs and a post-doc at the University of Michigan. He completed his Ph.D. at the Computer Science and Automation (CSA) department of the Indian Institute of Science (IISc), Bangalore advised by Prof. Shivani Agarwal.
5.15 p.m.–5.30 p.m.
BREAK
5.30 p.m.–6.00 p.m.
AbstractHistorically, Artificial Intelligence has taken a symbolic route for representing and reasoning about objects at a higher-level or a statistical route for learning complex models from large data. To achieve true AI, it is necessary to make these different paths meet and enable seamless human interaction. First, I briefly will introduce learning from rich, structured, complex and noisy data. Next, I will present the recent progress that allows for more reasonable human interaction where the human input is taken as “advice” and the learning algorithm combines this advice with data. The advice can be in the form of qualitative influences, preferences over labels/actions, privileged information obtained during training or simple precision-recall trade-off. Finally, I will outline our recent work on "closing-the-loop" where information is solicited from humans as needed that allows for seamless interactions with the human expert.
BioDr. Sriraam Natarajan is an Associate Professor and the Director for Center for ML at the Department of Computer Science at University of Texas Dallas. He was previously an Associate Professor and earlier an Assistant Professor at Indiana University, Wake Forest School of Medicine, a post-doctoral research associate at University of Wisconsin-Madison and had graduated with his PhD from Oregon State University. His research interests lie in the field of Artificial Intelligence, with emphasis on Machine Learning, Statistical Relational Learning and AI, Reinforcement Learning, Graphical Models and Biomedical Applications. He has received the Young Investigator award from US Army Research Office, Amazon Faculty Research Award, Intel Faculty Award, XEROX Faculty Award, Verisk Faculty Award and the IU trustees Teaching Award from Indiana University. He is the program co-chair of SDM 2020 and ACM CoDS-COMAD 2020 conferences. He is the specialty chief editor of Frontiers in ML and AI journal, an editorial board member of MLJ, JAIR and DAMI journals and is the electronics publishing editor of JAIR.
BioDr. Sriraam Natarajan is an Associate Professor and the Director for Center for ML at the Department of Computer Science at University of Texas Dallas. He was previously an Associate Professor and earlier an Assistant Professor at Indiana University, Wake Forest School of Medicine, a post-doctoral research associate at University of Wisconsin-Madison and had graduated with his PhD from Oregon State University. His research interests lie in the field of Artificial Intelligence, with emphasis on Machine Learning, Statistical Relational Learning and AI, Reinforcement Learning, Graphical Models and Biomedical Applications. He has received the Young Investigator award from US Army Research Office, Amazon Faculty Research Award, Intel Faculty Award, XEROX Faculty Award, Verisk Faculty Award and the IU trustees Teaching Award from Indiana University. He is the program co-chair of SDM 2020 and ACM CoDS-COMAD 2020 conferences. He is the specialty chief editor of Frontiers in ML and AI journal, an editorial board member of MLJ, JAIR and DAMI journals and is the electronics publishing editor of JAIR.
6.00 p.m.–6.30 p.m.
Srinivasan Parthasarathy
Abstract:Hurricane-induced flooding can lead to substantial loss of life and huge damage to infrastructure. Mapping flood extent from satellite or aerial imagery is essential for prioritizing relief efforts and for assessing future flood risk. Identification of water extent in such images can be challenging considering the heterogeneity in waterbody size and shape, cloud cover, and natural variations in landcover. In this effort, we introduce a novel cognitive framework basedon a semi-supervised learning algorithm, called HUman-GuidedFlood Mapping (HUG-FM), specifically designed to tackle the floodmapping problem.We test the efficacy and efficiency of our framework on imagery from several recent flood-induced emergencies and results show that our algorithm can robustly and correctly detect water logged areas compared to the state-of-the-art. We then evaluate whether expert guidance can be replaced by the wisdom of a crowd (e.g.,crisis volunteers). We design an online crowdsourcing platform based on HUG-FM and propose a novel ensemble method to lever-age crowdsourcing efforts. We conduct an experiment with over 50 participants and show that crowdsourced HUG-FM(CHUG-FM) can approach or even exceed the performance of a single expert providing guidance (HUG-FM).
6.30 p.m.–7 p.m.
Abstract : Our minds make inferences that appear to go far beyond standard machine learning. Whereas people can learn richer representations and use them for a wider range of learning tasks, machine learning algorithms have been mainly employed in a stand-alone context, constructing a single function from a table of training examples. In this talk, I shall touch upon a view on machine learning, called probabilistic programming, that can help capturing these human learning aspects by combining high-level programming languages and probabilistic machine learning — the high-level language helps reducing the cost of modelling and probabilities help quantifying when a machine does not know something. Since probabilistic inference remains intractable, existing approaches leverage deep learning for inference. Instead of “going down the full neural road,” I shall argue to use sum-product networks, a deep but tractable architecture for probability distributions. This can speed up inference in probabilistic programs, as I shall illustrate for unsupervised science understanding, and even pave the way towards automating density estimation, making machine learning accessible to a broader audience of non-experts. This talk is based on joint works with many people such as Carsten Binnig, Zoubin Ghahramani, Andreas Koch, Alejandro Molina, Sriraam Natarajan, Robert Peharz, Constantin Rothkopf, Thomas Schneider, Patrick Schramwoski, Xiaoting Shao, Karl Stelzner, Martin Trapp, Isabel Valera, Antonio Vergari, and Fabrizio Ventola.
Bio : Kristian Kersting is a full professor (W3) for AI and ML at TU Darmstadt. After receiving his Ph.D. from U. Freiburg in 2006, he was with MIT, Fraunhofer IAIS, U. Bonn, and TU Dortmund. His main research interests are (deep) probabilistic programming and learning. Kristian has published over 170 peer-reviewed articles. He is an EurAI Fellow, an ELLIS Fellow and received the inaugural German AI Award (Deutscher KI-Preis) 2019, as well as several paper awards (TPM 2019, AIIDE 2015, ECML 2006) and the EurAI Dissertation Award 2006. Kristian has been on the (senior) PC of major AI/ML conferences (e.g. AAAI, ICML, IJCAI, NeurIPS, ICLR, and CVPR) and co-chaired the PC of ECML PKDD.
Bio : Kristian Kersting is a full professor (W3) for AI and ML at TU Darmstadt. After receiving his Ph.D. from U. Freiburg in 2006, he was with MIT, Fraunhofer IAIS, U. Bonn, and TU Dortmund. His main research interests are (deep) probabilistic programming and learning. Kristian has published over 170 peer-reviewed articles. He is an EurAI Fellow, an ELLIS Fellow and received the inaugural German AI Award (Deutscher KI-Preis) 2019, as well as several paper awards (TPM 2019, AIIDE 2015, ECML 2006) and the EurAI Dissertation Award 2006. Kristian has been on the (senior) PC of major AI/ML conferences (e.g. AAAI, ICML, IJCAI, NeurIPS, ICLR, and CVPR) and co-chaired the PC of ECML PKDD.
June 19
June 19
2.00 p.m.–2.30 p.m.
Abstract:Two of the most challenging problems in epilepsy are (1) the ability to objectively assess patient outcomes, and (2) quickly arriving at optimal therapy.
The NeuroPace RNS System is a closed-loop brain responsive neurostimulator that is FDA approved for the adjunctive treatment of patients with medically intractable partial onset epilepsy having 1-2 seizure foci. The RNS System also captures snippets of brain activity in electrocorticographic (ECoG) files. Till date, over 5 million ECoG files have been captured from patients implanted with the RNS System. This large dataset of brain activity recordings facilitates the development of machine learning algorithms for assessing patient outcomes and proposing potential stimulation settings.
Bio:Sharanya Desai received B.E in EEE and M.Sc in Biological Sciences from BITS-Pilani, India and PhD in Bioengineering from Georgia Tech, USA. During her graduate studies in Dr. Steve Potter's lab at Georgia Tech and Dr. Robert Gross's lab at Emory University, she worked on a novel multielectrode microstimulation approach for reducing seizures in a rodent model of epilepsy. She currently works as a Principal Research Scientist at NeuroPace - a medical device company that manufactures closed-loop brain responsive neurostimulators for epilepsy. Her research focus is on applying data science techniques for understanding and improving chronic brain diseases.
Sharanya is a recipient of the International Schlumberger Faculty for the Future fellowship and author of publications and patents with a focus on understanding and improving human health.
The NeuroPace RNS System is a closed-loop brain responsive neurostimulator that is FDA approved for the adjunctive treatment of patients with medically intractable partial onset epilepsy having 1-2 seizure foci. The RNS System also captures snippets of brain activity in electrocorticographic (ECoG) files. Till date, over 5 million ECoG files have been captured from patients implanted with the RNS System. This large dataset of brain activity recordings facilitates the development of machine learning algorithms for assessing patient outcomes and proposing potential stimulation settings.
Bio:Sharanya Desai received B.E in EEE and M.Sc in Biological Sciences from BITS-Pilani, India and PhD in Bioengineering from Georgia Tech, USA. During her graduate studies in Dr. Steve Potter's lab at Georgia Tech and Dr. Robert Gross's lab at Emory University, she worked on a novel multielectrode microstimulation approach for reducing seizures in a rodent model of epilepsy. She currently works as a Principal Research Scientist at NeuroPace - a medical device company that manufactures closed-loop brain responsive neurostimulators for epilepsy. Her research focus is on applying data science techniques for understanding and improving chronic brain diseases.
Sharanya is a recipient of the International Schlumberger Faculty for the Future fellowship and author of publications and patents with a focus on understanding and improving human health.
2.30 p.m.–3.00 p.m.
Helping machine learning to help us in personalized medicine
Helping machine learning to help us in personalized medicine
Julio Saez-Rodriguez
Abstract:One area where artificial intelligence is expected to have a major impact in the health area is by developing algorithms that help us provide the right drug for each patient, that is, for personalized medicine. In this talk I will discuss our work applying machine learning on large pharmaco-genomic screenings in cell lines to build predictive models. Integration of this data with prior knowledge on signaling pathways and transcription factors provides biomarkers and offer hypotheses for novel combination therapies. Our own analysis as well as the results of a crowdsourcing effort (as part of a DREAM challenge) reveals that prediction of drug efficacy is far from accurate, implying important limitations for personalised medicine. An important aspect that deserves further attention is the dynamics of signaling networks and how they response to perturbations such as drug treatment. I will present how cell-specific logic models, trained with measurements upon perturbations, can provides new biomarkers and treatment opportunities not noticeable by static molecular characterisation. In summary, I will advocate that combining the right data with biological knowledge will be important to build predictive models for personalized medicine.
Bio:Julio is Professor of Medical Bioinformatics and Data Analysis, University of Heidelberg, and director of the Institute of computational biomedicine. He is also a group leader of the EMBL-Heidelberg University Molecular Medicine Partnership Unit, and a co-director of the DREAM challenges (http://dreamchallenges.org) to crowdsource computational systems biology. He obtained his M.S. in Chemical Engineering in 2001, and a PhD in 2007 at the University of Magdeburg and the Max-Planck-Institute. He was a postdoctoral fellow at Harvard Medical School and M.I.T., and a Scientific Coordinator of the NIH-NIGMS Cell Decision Process Center from 2007 to 2010. From 2010 until 2015 he was a group leader at EMBL-EBI with a joint appointment in the EMBL Genome Biology Unit in Heidelberg, as well as a senior fellow at Wolfson College (Cambridge). From 2015 to 2018 he was professor of Computational Biomedicine at the RWTH University Medical Hospital in Aachen, Germany.He is interested in developing and applying computational methods to acquire a functional understanding of signaling networks and their deregulation in disease, and to apply this knowledge to develop novel therapeutics. Current emphasis in his group is on use of single-cell technologies, multi-omics integration, and understanding multi-cellular communication.
Bio:Julio is Professor of Medical Bioinformatics and Data Analysis, University of Heidelberg, and director of the Institute of computational biomedicine. He is also a group leader of the EMBL-Heidelberg University Molecular Medicine Partnership Unit, and a co-director of the DREAM challenges (http://dreamchallenges.org) to crowdsource computational systems biology. He obtained his M.S. in Chemical Engineering in 2001, and a PhD in 2007 at the University of Magdeburg and the Max-Planck-Institute. He was a postdoctoral fellow at Harvard Medical School and M.I.T., and a Scientific Coordinator of the NIH-NIGMS Cell Decision Process Center from 2007 to 2010. From 2010 until 2015 he was a group leader at EMBL-EBI with a joint appointment in the EMBL Genome Biology Unit in Heidelberg, as well as a senior fellow at Wolfson College (Cambridge). From 2015 to 2018 he was professor of Computational Biomedicine at the RWTH University Medical Hospital in Aachen, Germany.He is interested in developing and applying computational methods to acquire a functional understanding of signaling networks and their deregulation in disease, and to apply this knowledge to develop novel therapeutics. Current emphasis in his group is on use of single-cell technologies, multi-omics integration, and understanding multi-cellular communication.
3.00 p.m.–3.30 p.m.
Predicting disease susceptibility and disease spread in H1N1 influenza
Predicting disease susceptibility and disease spread in H1N1 influenza
Nagasuma Chandra
Abstract:Despite extensive work on influenza, a number of questions still remain open about why individuals are differently susceptible to the disease and why only some strains lead to epidemics. We study the effect of human leukocyte antigen (HLA) genotype heterogeneity on possible cytotoxic T-lymphocyte (CTL) response to 186 influenza H1N1 genomes. To enable such analysis, we first systematically characterize the variations in HLA alleles and their peptide recognition properties and develop a new grouping scheme based on comparison of binding-site structures, which provides an insightful classification and also rationalizes the physicochemical basis of recognition specificity. This has also led to the development of a sensitive predictor of HLA epitopes from genome sequences, using which we obtain a comprehensive list of possible high-affinity epitopes from H1N1 influenza genomes. We then reconstruct HLA genotypes in different populations using allele frequency data in different populations and a probabilistic method and use the HLA genotype distribution to assess the pool of epitopes that individual hosts can recognize from each of the viral strains. We find that large populations can be classified into a small number of groups called response-types, specific to a given viral strain, providing a basis to predict disease susceptibility. We then use this information to model disease spread using a variation of the standard SIR model so as to incorporate genetic information. Our results show that HLA allele profiles which lead to a large spread in individual susceptibility values can act as a protective barrier against the spread of influenza. We predict that populations skewed such that a small number of highly susceptible individuals coexist with a large number of less susceptible ones, should exhibit smaller outbreaks than populations with the same average susceptibility but distributed more uniformly across individuals. Overall, the work presents a novel conceptual framework towards understanding how genetic heterogeneity influences disease susceptibility in individuals and in populations.
Bio:Nagasuma Chandra is a molecular systems biologist and bioinformatician and works on modelling complex biological processes and applying them to study human health and disease. Her research is interdisciplinary involving computational modelling of complex biological processes to address fundamental questions about how genome-wide molecular networks respond to a variety of pathological conditions and how that knowledge can be translated into biomedical applications. She is currently a Professor at the Department of Biochemistry, Indian Institute of Science and additionally affiliated with Bioengineering and Mathematical biology initiatives at the Institute. She received her PhD from the University of Bristol, UK. She is an elected fellow of the Indian Academy of Sciences.
Bio:Nagasuma Chandra is a molecular systems biologist and bioinformatician and works on modelling complex biological processes and applying them to study human health and disease. Her research is interdisciplinary involving computational modelling of complex biological processes to address fundamental questions about how genome-wide molecular networks respond to a variety of pathological conditions and how that knowledge can be translated into biomedical applications. She is currently a Professor at the Department of Biochemistry, Indian Institute of Science and additionally affiliated with Bioengineering and Mathematical biology initiatives at the Institute. She received her PhD from the University of Bristol, UK. She is an elected fellow of the Indian Academy of Sciences.
3.30 p.m.–3.45 p.m.
BREAK
3.45 p.m.–4.15 p.m.
From Genomes To Systems-Level Models- Towards leveraging reprogrammed metabolism In Cancer And Infectious Disease
From Genomes To Systems-Level Models- Towards leveraging reprogrammed metabolism In Cancer And Infectious Disease
Anu Raghunathan
Abstract:Intrinsic to systems biology (the holistic study of interacting genes and gene products of an organism) is the notion that biological systems have “emergent properties”. Such biological outcomes cannot be predicted by traditional reductionist methods for one organism and get increasingly complex and intractable for two interacting systems. Any organism can be represented qualitatively as a genome-scale biochemical reaction network reconstruction. Such reconstructions are essentially stoichiometric representations of reactions that define cellular function and can be converted into a mathematical format to compute cell phenotype. Further, use of a constraints-based modeling philosophy allows simulating the physiological function of the cell's complex networks and interactions. The ability to compute and predict cell function allows us diverse applications ranging from designing and engineering cells for biotechnology to understanding pathogenesis in medicine. In this talk, I will discuss the paradigm of metabolic systems biology. I will give an overview of the work done in our laboratory to not only build such models but also their application in two major areas of medicine- Infectious Disease and Cancer. The use of such models to understand the emergent phenomenon of drug resistance from a metabolic standpoint will be discussed. Integrated with clinical data, such systems-approaches present scalable personalized and individualized therapeutic solutions for human health.
Bio:Anu Raghunathan is currently a Senior Principal Scientist in the Chemical Engineering Division at CSIR-National Chemical Laboratory in Pune. She has been with NCL for the last 10 years. Before that she was working as a faculty in Infectious Diseases at Mount Sinai School of Medicine in NY, NY. Her research group, the Metabolic Inquiry and Cellular Engineering (MICE Lab) group, uses systems approaches (both experimental and computational) to understand biological cell behavior and function. She works on applying these principles to understand drug/antibiotic resistance in cells and also manipulate cells to make desired products. She is an associate editor for Biosystems an Elsevier Publication and an active board member of the International Study Group for Systems Biology.
Bio:Anu Raghunathan is currently a Senior Principal Scientist in the Chemical Engineering Division at CSIR-National Chemical Laboratory in Pune. She has been with NCL for the last 10 years. Before that she was working as a faculty in Infectious Diseases at Mount Sinai School of Medicine in NY, NY. Her research group, the Metabolic Inquiry and Cellular Engineering (MICE Lab) group, uses systems approaches (both experimental and computational) to understand biological cell behavior and function. She works on applying these principles to understand drug/antibiotic resistance in cells and also manipulate cells to make desired products. She is an associate editor for Biosystems an Elsevier Publication and an active board member of the International Study Group for Systems Biology.
4.15 p.m.–4.45 p.m.
Abstract:Preterm birth is conventionally defined as a birth that occurs before 37 completed weeks ofgestation. It is a unique disease in the way it is determined by the duration of gestation andnot by a pathological process. The duration of gestation is the period between the date ofconception and date of delivery. While the date of birth can be documented with fairaccuracy, ascertaining the date of conception is challenging. The estimation of gestationalage during pregnancy also called as the dating of pregnancy has been conventionally doneusing the first day of the recall-based last menstrual period or measurement of foetalbiometry by ultrasound. Each of these methods poses a unique set of challenges. Differentformulae have been developed globally for estimation of gestational age byultrasonography in the first and second trimester of pregnancy. In this talk, I will talk aboutour efforts to build an Indian population-specific dating formula and compare itsperformance with published formulae. Finally, I will discuss the implications of the choice ofdating method on the preterm birth rate. The data for this study was from GARBH-Ini, anongoing pregnancy cohort of North Indian women to study preterm birth.
Bio:Himanshu Sinha is currently an Associate Professor at the Department of Biotechnology,Bhupat and Jyoti Mehta School of Biosciences, IIT Madras and coordinates Initiative forBiological Systems Engineering (IBSE), an interdisciplinary centre to study biological systemsand develop innovative approaches and algorithms for analysis of clinical and healthcaredata. He received his PhD from University of Cambridge, UK, in 2000; followed by postdocstints at Duke University Medical Center, Durham, NC USA and European Molecular BiologyLaboratory (EMBL), Heidelberg, Germany. At IIT Madras, he leads Systems Genetics Lab,where his group works to uncover complex relationships between genetic and phenotypicvariation and the role of evolution in changing the dynamics of these relationships; andapply computational methods on various omics and clinical datasets to model relationshipsbetween genetic variation and disease susceptibility.
Bio:Himanshu Sinha is currently an Associate Professor at the Department of Biotechnology,Bhupat and Jyoti Mehta School of Biosciences, IIT Madras and coordinates Initiative forBiological Systems Engineering (IBSE), an interdisciplinary centre to study biological systemsand develop innovative approaches and algorithms for analysis of clinical and healthcaredata. He received his PhD from University of Cambridge, UK, in 2000; followed by postdocstints at Duke University Medical Center, Durham, NC USA and European Molecular BiologyLaboratory (EMBL), Heidelberg, Germany. At IIT Madras, he leads Systems Genetics Lab,where his group works to uncover complex relationships between genetic and phenotypicvariation and the role of evolution in changing the dynamics of these relationships; andapply computational methods on various omics and clinical datasets to model relationshipsbetween genetic variation and disease susceptibility.
4.45 p.m.– 5.15 p.m.
Abstract:Constraint-based metabolic modeling serves as the gold standard for interrogating thegenotype-phenotype relationship, typically in healthy v/s disease conditions. The mostcomprehensive human metabolic network, Recon 3D (Brunk et al, 2018, Nat. Biotechnol.)marks as the starting point for a number of biomedical applications. This catalogues thehuman-specific metabolic reactions, along with their genetic and protein information. Whenused to derive an organ-specific sub-network, Recon 3D has tremendous potential for gainingmechanistic insights for diseases. The most recent, whole-body metabolic networks, deemedHarvey and Harvetta (Thiele et al, 2020, Mol. Syst. Biol.) generated thereof, is a breakthroughfor the envisioned ‘’virtual human’’. Application of these models with respect to capturingdisease-specific biomarkers, representing known inter-organ metabolic cycles, and derivinggut microbiome-specific personalized models will be discussed.
Bio:Swagatika Sahoo is currently DST-INPIRE Faculty at the Dept. of Chemical Engineering,and core faculty at the Initiative for Biological Systems Engineering, IIT Madras. Shereceived her Ph.D. from the Centre for Systems Biology, University of Iceland in 2014,followed by her post doc from the Luxembourg Centre for Systems Biomedicine, Universityof Luxembourg. Since, 2016 she is with IIT Madras, where in, her research lab focuses onusing systems biology approaches for predicting therapeutic strategies for metabolic diseases,including brain disorders and cancer. She has served reviewer for Bioinformatics (OxfordAcademic) & Journal of Inherited Metabolic Disease (Official Journal of the Society for theStudy of Inborn Errors of Metabolism).
Bio:Swagatika Sahoo is currently DST-INPIRE Faculty at the Dept. of Chemical Engineering,and core faculty at the Initiative for Biological Systems Engineering, IIT Madras. Shereceived her Ph.D. from the Centre for Systems Biology, University of Iceland in 2014,followed by her post doc from the Luxembourg Centre for Systems Biomedicine, Universityof Luxembourg. Since, 2016 she is with IIT Madras, where in, her research lab focuses onusing systems biology approaches for predicting therapeutic strategies for metabolic diseases,including brain disorders and cancer. She has served reviewer for Bioinformatics (OxfordAcademic) & Journal of Inherited Metabolic Disease (Official Journal of the Society for theStudy of Inborn Errors of Metabolism).
5.15 p.m.– 5.30 p.m.
BREAK
5.30 p.m.–6.00 p.m.
Can cellphone data help fight Covid-19 in India?
Can cellphone data help fight Covid-19 in India?
Satchit Balsari
Abstract:The spread of Covid19 across the world was accompanied by a near-logarithmic growth of technological solutions deployed to track, trace, predict and curtail the pandemic. Leveraging data from GPS traces and bluetooth signals, companies around the world began delivering tools to governments, and in some cases directly to the public, that collated data from millions of users around the world, forever changing the crisis response landscape. Mobility data and algorithms are expected to be part of our built environment going forward, with profound implications on privacy. This talk shares early observations from the Covid-19 Mobility Data Network, a research intermediary of over 70 universities that works with public health agencies across the world.
Bio:Dr. Satchit Balsari is assistant professor at the medical and public health schools at Harvard, where he directs the India Digital Health Network, at the Lakshmi Mittal South Asia Institute.His interdisciplinary interests in digital technology, disaster response, and population health have resulted in innovative applications of mobile, cloud-based technology to address public health challenges in mass gatherings, disasters, and humanitarian crises. Balsari’s signature initiatives include project EMcounter (a customizable, portable digital surveillance tool, the latest iteration of which was used at the world’s largest mass gathering, the Kumbh Mela in India) and Voices, a crowd-sourced, online disaster response analysis tool. He co-founded the Covid19 Mobility Data Network, in March 2020.
Bio:Dr. Satchit Balsari is assistant professor at the medical and public health schools at Harvard, where he directs the India Digital Health Network, at the Lakshmi Mittal South Asia Institute.His interdisciplinary interests in digital technology, disaster response, and population health have resulted in innovative applications of mobile, cloud-based technology to address public health challenges in mass gatherings, disasters, and humanitarian crises. Balsari’s signature initiatives include project EMcounter (a customizable, portable digital surveillance tool, the latest iteration of which was used at the world’s largest mass gathering, the Kumbh Mela in India) and Voices, a crowd-sourced, online disaster response analysis tool. He co-founded the Covid19 Mobility Data Network, in March 2020.
6.00 p.m.–6.30 p.m.
Abstract:Drug development has become an economically unsustainable process due to low reproducibility, high attrition rates, failures in Phase III trials, and increasing R&D costs. These trends have necessitated the modeling of human diseases as networks to simplify complex multi-cellular processes, understand patterns in noisy data that humans cannot find, and thereby, achieve precision. Using Inflammatory Bowel Disease (IBD) as an example, here we outline an AI-assisted approach for target identification and validation. We built a network in which clusters of genes are connected by directed edges that highlight asymmetric Boolean relationships. Algorithms of machine learning were used to sift through the network to pinpoint the path of continuum states that most reliably distinguished between health and disease states, disease severity, and predicted treatment outcome. Our network-based computational models reliably classified healthy and IBD-afflicted tissues in several publicly available gene expression datasets (906 human samples, 234 mouse samples). The AI-identified path was enriched in gene clusters that maintain the integrity of the gut epithelial barrier. We exploit the gene clusters on that path for prioritizing one target, choosing appropriate pre-clinical murine models for target validation and for designing patient-derived organoid models. Treatment efficacy is confirmed in these patient-models using a multivariate analysis. This AI-assisted approach also predicts Phase III success in IBD with higher accuracy over traditional approaches. The combined synergy of AI-assisted target identification and the choice of preclinical mouse and human models for target validation has provided a first-in-class gut barrier-protective agent in IBD.
Bio:Debashis Sahoo received B.Tech. degree in Computer Science and Engineering from Indian Institute of Technology, Kharagpur, in 2000, and the M.S. and Ph.D. degrees in Electrical Engineering from the Stanford University, in 2003 and 2008, respectively. He is currently a joint Assistant Professor of Computer Science and Engineering and Pediatrics at University of California, San Diego. His research interests include analysis of biological systems and formal verification of hardware designs. Dr. Sahoo pioneered identifying simple Boolean relationships between gene expression values. Dr. Sahoo’s research has produced impacts in understanding several biological systems including, B cell differentiation, colon cancer, bladder cancer, and prostate cancer. Among the successes was Dr. Sahoo’s previous work where he identified the expression level of CDX2 as a predictive biomarker for favorable response to conventional chemotherapy among stage II colon cancer patients (Dalerba, Sahoo et al. NEJM 2016). Dr. Sahoo has received many grants for his achievements, notable among them are the prestigious NIH Pathway to Independence K99/R00 Award, DOD Prostate Cancer Training Award, and Bladder Cancer Advocacy Network Young Investigator Award. He has continued his status as a highly accomplished researcher, publishing in high-impact journals and receiving grants for his computational biology adventure.
Bio:Debashis Sahoo received B.Tech. degree in Computer Science and Engineering from Indian Institute of Technology, Kharagpur, in 2000, and the M.S. and Ph.D. degrees in Electrical Engineering from the Stanford University, in 2003 and 2008, respectively. He is currently a joint Assistant Professor of Computer Science and Engineering and Pediatrics at University of California, San Diego. His research interests include analysis of biological systems and formal verification of hardware designs. Dr. Sahoo pioneered identifying simple Boolean relationships between gene expression values. Dr. Sahoo’s research has produced impacts in understanding several biological systems including, B cell differentiation, colon cancer, bladder cancer, and prostate cancer. Among the successes was Dr. Sahoo’s previous work where he identified the expression level of CDX2 as a predictive biomarker for favorable response to conventional chemotherapy among stage II colon cancer patients (Dalerba, Sahoo et al. NEJM 2016). Dr. Sahoo has received many grants for his achievements, notable among them are the prestigious NIH Pathway to Independence K99/R00 Award, DOD Prostate Cancer Training Award, and Bladder Cancer Advocacy Network Young Investigator Award. He has continued his status as a highly accomplished researcher, publishing in high-impact journals and receiving grants for his computational biology adventure.
June 20
June 20
2.00 p.m.–2.30 p.m.
Integration of Machine Learning with Domain Knowledge for Manufacturing Applications Video recording
Integration of Machine Learning with Domain Knowledge for Manufacturing Applications Video recording
Raghunathan Rengasamy
2.30 p.m.–3.00 p.m.
Abstract: Object detection has been one of the fundamental problems in computer vision which aims to detect and identify the location of objects of a particular class such as humans, cars, etc., in an image. It forms the basis of many vision tasks such as surveillance, image captioning, object tracking, etc. With the development of various deep learning models, much research attention has been devoted to object detection, leading to several significant improvements in terms of architecture and inference time. Most of the object detection research in the past few years has been on natural images with real-life objects. The goal of this work is to study object detection for a very different class of images, namely computer-generated scientific plots. Scientific plots such as bar plots, line plots, etc. provide an efficient way of visually representing the data where a table cannot adequately demonstrate the meaningful relationships or patterns between data points. Such plots are frequently found in textbooks, technical reports, academic papers, etc. Interpreting and understanding the underlying data encoded in these plots is considered a test of human aptitude. Hence, it is of interest to build systems which can understand and reason over scientific plots.
Scientific plots differ from natural images in three crucial ways. First, unlike natural images, they combine both text (e.g., axes, tick labels) and visual elements (e.g., bars, and legends). Second, they exhibit significant variation in scale and aspect ratio of objects (e.g., thin dot-lines and thick, long bars). Lastly, there are underlying structural relationships between objects (e.g., a tick label and corresponding bar in a bar-plot) which can be exploited for better understanding and reasoning. Further, localization accuracy is significantly more critical for plots than for natural images. This leads to the following interesting question, “Are existing object detection methods adequate for detecting text and visual elements in scientific plots which are arguably different from the objects found in natural images?” To answer this question, we train and compare the accuracy of nine state-of-the-art object detection networks on the PlotQA dataset with over 220,000 scientific plots. At the standard IOU setting of 0.5, most networks perform well with mAP scores higher than 80% in detecting the relatively simple objects in plots. However, the performance drops drastically when evaluated at a stricter IOU of 0.9 with the best model giving an mAP of 35.70%. Note that such a stricter evaluation is essential when dealing with scientific plots where even minor localization errors can lead to significant fallacies in downstream numerical inferences.
Given this poor performance, we propose minor modifications to existing models by combining ideas from different object detection networks. While this significantly improves the performance, there are still two main issues: (i) performance on textual objects which are essential for reasoning is abysmal, and (ii) inference time is unacceptably large considering the simplicity of plots. Based on these experiments and results, we identify the following considerations for improving object detection on plots: (a) lower inference time, (b) higher precision on textual objects, and (c) more accurate localization with a custom loss function with non-negligible loss values at high IOU (> 0.8). We propose a network, namely, PlotNet which meets all these considerations: It is 16x faster than the best performing competitor and significantly improves upon the accuracy of existing models with an mAP of 93.44% at an IOU of 0.9.
Bio:Mitesh M. Khapra is an Assistant Professor in the Department of Computer Science and Engineering at IIT Madras. While at IIT Madras he plans to pursue his interests in the areas of Deep Learning, Multimodal Multilingual Processing, Dialog systems and Question Answering. Prior to that he worked as a Researcher at IBM Research India. During the four and half years that he spent at IBM he worked on several interesting problems in the areas of Statistical Machine Translation, Cross Language Learning, Multimodal Learning, Argument Mining and Deep Learning. This work led to publications in top conferences in the areas of Computational Linguistics and Machine Learning. Prior to IBM, he completed his PhD and M.Tech from IIT Bombay in Jan 2012 and July 2008 respectively. His PhD thesis dealt with the important problem of reusing resources for multilingual computation. During his PhD he was a recipient of the IBM PhD Fellowship and the Microsoft Rising Star Award. He is also a recipient of the Google Faculty Research Award, 2018.
Scientific plots differ from natural images in three crucial ways. First, unlike natural images, they combine both text (e.g., axes, tick labels) and visual elements (e.g., bars, and legends). Second, they exhibit significant variation in scale and aspect ratio of objects (e.g., thin dot-lines and thick, long bars). Lastly, there are underlying structural relationships between objects (e.g., a tick label and corresponding bar in a bar-plot) which can be exploited for better understanding and reasoning. Further, localization accuracy is significantly more critical for plots than for natural images. This leads to the following interesting question, “Are existing object detection methods adequate for detecting text and visual elements in scientific plots which are arguably different from the objects found in natural images?” To answer this question, we train and compare the accuracy of nine state-of-the-art object detection networks on the PlotQA dataset with over 220,000 scientific plots. At the standard IOU setting of 0.5, most networks perform well with mAP scores higher than 80% in detecting the relatively simple objects in plots. However, the performance drops drastically when evaluated at a stricter IOU of 0.9 with the best model giving an mAP of 35.70%. Note that such a stricter evaluation is essential when dealing with scientific plots where even minor localization errors can lead to significant fallacies in downstream numerical inferences.
Given this poor performance, we propose minor modifications to existing models by combining ideas from different object detection networks. While this significantly improves the performance, there are still two main issues: (i) performance on textual objects which are essential for reasoning is abysmal, and (ii) inference time is unacceptably large considering the simplicity of plots. Based on these experiments and results, we identify the following considerations for improving object detection on plots: (a) lower inference time, (b) higher precision on textual objects, and (c) more accurate localization with a custom loss function with non-negligible loss values at high IOU (> 0.8). We propose a network, namely, PlotNet which meets all these considerations: It is 16x faster than the best performing competitor and significantly improves upon the accuracy of existing models with an mAP of 93.44% at an IOU of 0.9.
Bio:Mitesh M. Khapra is an Assistant Professor in the Department of Computer Science and Engineering at IIT Madras. While at IIT Madras he plans to pursue his interests in the areas of Deep Learning, Multimodal Multilingual Processing, Dialog systems and Question Answering. Prior to that he worked as a Researcher at IBM Research India. During the four and half years that he spent at IBM he worked on several interesting problems in the areas of Statistical Machine Translation, Cross Language Learning, Multimodal Learning, Argument Mining and Deep Learning. This work led to publications in top conferences in the areas of Computational Linguistics and Machine Learning. Prior to IBM, he completed his PhD and M.Tech from IIT Bombay in Jan 2012 and July 2008 respectively. His PhD thesis dealt with the important problem of reusing resources for multilingual computation. During his PhD he was a recipient of the IBM PhD Fellowship and the Microsoft Rising Star Award. He is also a recipient of the Google Faculty Research Award, 2018.
3.00 p.m.–3.30 p.m.
Abstract:This talk outlines the challenges of E-Mobility in India, and how the unique context of Cleantech, Fintech and Digital Innovation will lead to interesting and world-leading solutions in our market. India can lead by showing the world how e-mobility services can be offered affordably to a billion people. The upfront cost challenges can be solved via a combination of asset sizing (2/3 wheelers and shared 4-wheelers), As-a-service business models (subscriptions, fleets, shared assets, micromobility & seamless multi-modal services), and digital (via cloud/AI/ML & IOT) linkages between assets and finance/insurance. The growth of renewables (especially solar) combined with storage applied in decentralized contexts will also allow enterprises and communities to take leadership in driving this revolution aided by fintech innovation to solve the problem of high upfront capital costs, and resilience in a weak utility grid context. E-Commerce and logistics will also play an important role in adopting and driving e-Mobility.Ref: https://www.linkedin.com/pulse/e-mobility-cleantech-fintech-digital-nexus-india-kalyanaraman/
Bio:Shiv is CTO, Energy & Mobility, Microsoft R&D India for Azure Global. Previously Shiv was Executive General Manager of Growth Offerings at GE Power Conversion responsible for new offerings in e-Mobility & digital/AI. Earlier he was at IBM Research - India, and the Chief Scientist of IBM Research - Australia. Before IBM, he was a tenured Full Professor at Rensselaer Polytechnic Institute in Troy, NY, USA. He has degrees from Indian Institute of Technology, Madras (B.Tech), Ohio State University (MS, PhD) and RPI (MBA). Shiv is a Fellow of the IEEE (2010), Fellow of Indian National Academy of Engineering (2015), ACM Distinguished Scientist (2010), MIT Technology Review TR100 young innovator (1999).
Bio:Shiv is CTO, Energy & Mobility, Microsoft R&D India for Azure Global. Previously Shiv was Executive General Manager of Growth Offerings at GE Power Conversion responsible for new offerings in e-Mobility & digital/AI. Earlier he was at IBM Research - India, and the Chief Scientist of IBM Research - Australia. Before IBM, he was a tenured Full Professor at Rensselaer Polytechnic Institute in Troy, NY, USA. He has degrees from Indian Institute of Technology, Madras (B.Tech), Ohio State University (MS, PhD) and RPI (MBA). Shiv is a Fellow of the IEEE (2010), Fellow of Indian National Academy of Engineering (2015), ACM Distinguished Scientist (2010), MIT Technology Review TR100 young innovator (1999).
3.30 p.m.–3.45 p.m.
BREAK
3.45 p.m.–4.15 p.m.
Bio:Dr. Niloy Ganguly is a Professor in the Dept. of Computer Science and Engineering at IIT Kharagpur and a Fellow of Indian Academy of Engineering. He spent 2 years as a Research Scientist in Technical University, Dresden, before joining IIT Kharagpur in 2005, and has risen to the rank of Professor in 2014. He has done his Btech from IIT Kharagpur and his Phd from IIEST, Shibpur. His research interests lie primarily in Social Computing, Machine Learning, and Network Science. He has published in 60 journals and 140 conferences, several of which are in reputed international venues such as NIPS, KDD, ICDM, IJCAI, WWW, CSCW, EMNLP, CHI, ICWSM, INFOCOM, Physical Reviews, IEEE and ACM Transaction etc. He has served in the program committee of COMSNETS, NetSciCom, JCDL, WWW, DEBS and CODS. Prof Ganguly’s work has been recognized through awards by NSF, Cisco, NetApp, Samsung, and Yahoo!, among others. He has received prestigious research grants and projects, notably from Data Transparency Lab, IMPRINT, ITRA, Intel, HPE, Adobe, Microsoft Research, Accenture, BEL, and TCS. He has guided 16 Ph.D. and 5 M.S. students during this tenure. He is the founding member of the Complex Networks Research Group (CNeRG), comprising faculty members, research scholars, and other students affiliated to the department. The group is a success story in itself, with several long-standing impactful collaborations, and presence in reputed venues across domains such as Social Computing, Machine Learning and Deep Learning, Natural Language Processing, Network Science, Networked Systems, etc.
4.15 p.m.–4.45 p.m.
Bio:Tal Raviv is an associate professor of Industrial Engineering at the Iby and Aladar Fleischman Faculty of Engineering, Tel Aviv University, Israel. He holds a BA from the Eitan Berglas School of Economics, Tel Aviv University (1993), an MBA from the Recanati School of Business, Tel Aviv University (1997), and a Ph.D. in Operations Research from the William Davidson Faculty of Industrial Engineering and Management, Technion - Israel Institute of Technology, Haifa (2003). He spent two years (2004-2006) as a postdoctoral fellow in the Sauder School of Business at the University of British Columbia, Vancouver, Canada. He published many papers in the operations research literature, served as an advisor for start-up companies, and as an editorial board member in Omega and Transportation Research Part B as well as a guest editor in the Euro Journal of Transportation and Logistics and Networks. His primary research interest is in transportation and logistics, with a focus on smart and sustainable transportation. He is currently studying topics such as shared mobility systems, small parcel delivery logistics, public transit planning, parking policy, and congestion fee. Tal is co-heading the transportation and logistics lab in the faculty of engineering and serves as the head of the Shlomo Shmeltzer Institute for smart transportation in Tel Aviv University.
4.45 p.m.–5.15 p.m.
Bayesian Matrix and Tensor Completion Methods and its Applications in Environmental Sensing and Transportation Video recording
Bayesian Matrix and Tensor Completion Methods and its Applications in Environmental Sensing and Transportation Video recording
Pravesh Biyani
Abstract: In this talk, we discuss two problems of importance - air quality monitoring and traffic estimation and prediction. We discuss the common theme that binds the two problems -- an incomplete matrix. We propose a moving sensor based air quality measurements and discuss how matrix completion can be used to obtain high fidelity dense air quality measurements. The Matrix completion and robust principal componentanalysis have been widely used for the recovery of data suffering from missing entries or outliers. In the above applications however,, the data is also time-varying, and the naive approach of per-snapshot recovery is both expensive and sub-optimal. In this talk, we will discuss generative Bayesian models that fit sequential multivariate measurements arising from a low-dimensional time-varying subspace. A variational Bayesian subspace filtering approach will be discussed that learns the underlying subspace and its state-transition matrix. Extensive tests over traffic and air quality data demonstrate the superior imputation, outlier rejection, and temporal prediction prowess of the proposed algorithm over the state-of-the-art matrix/tensorcompletion algorithms.
Bio:Pravesh is an associate professor at IIIT Delhi. He received his B.Tech from IIT Bombay in 2002 and MS from McMaster University in the year 2004. In late 2012, he was a post- doctoral researcher at the University of Minnesota, Minneapolis. He also won the INSPIRE Faculty award by the Govt. of India in 2012. His research interests lie in the intersection of signal processing and machine learning with applications in urban transportation and speech/audio processing. He is passionate about public transportation and is interested in building systems as well as policy research to improve urban mobility. He has specifically been an active supporter of the open transit data systems.
Bio:Pravesh is an associate professor at IIIT Delhi. He received his B.Tech from IIT Bombay in 2002 and MS from McMaster University in the year 2004. In late 2012, he was a post- doctoral researcher at the University of Minnesota, Minneapolis. He also won the INSPIRE Faculty award by the Govt. of India in 2012. His research interests lie in the intersection of signal processing and machine learning with applications in urban transportation and speech/audio processing. He is passionate about public transportation and is interested in building systems as well as policy research to improve urban mobility. He has specifically been an active supporter of the open transit data systems.
5:15 p.m.–5.30 p.m.
Break
5.30 p.m.–6.00 p.m.
Abstract:Like all verticals, AI has been making a game-changing impact on various facets of the Finance ecosystem. Going beyond the research bench, many predictive AI systems are enabling real-life 'Finance Applications' impacting lives of millions of people. In this session, I will talk about predictive modeling problems in Finance and the associated role of data with particular emphasis on evolving nature of it. The dual objectives being reducing the number of 'credit invisibles' as well as improving the accuracy of predictions. In the first part, the talk will be about the industry trends around how alternate data is emerging and the promises around them. The second part comprises a specific case study on an innovative technique measuring utility of alternate data and selectively influencing traditional data based models for improving the prediction performance.
Bio:Shourya Roy is the Head and Vice President of Amex AI Labs which is a center of excellence developing AI capabilities and solutions for addressing large scale real-life business problems. Prior to joining Amex in 2016, he spent nearly fifteen years in the labs of IBM and Xerox playing several leadership roles in technical research as well as research and strategic management. Shourya’s interest spans generally in machine learning and AI with specific focus on NLP as well as challenges pertaining to real-life development and deployment of AI techniques.
Bio:Shourya Roy is the Head and Vice President of Amex AI Labs which is a center of excellence developing AI capabilities and solutions for addressing large scale real-life business problems. Prior to joining Amex in 2016, he spent nearly fifteen years in the labs of IBM and Xerox playing several leadership roles in technical research as well as research and strategic management. Shourya’s interest spans generally in machine learning and AI with specific focus on NLP as well as challenges pertaining to real-life development and deployment of AI techniques.
6.00 p.m.–6.30 p.m.
AI for Public Health and Conservation: Learning and Planning in the Data-to-Deployment Pipeline Video recording
AI for Public Health and Conservation: Learning and Planning in the Data-to-Deployment Pipeline Video recording
Milind Tambe
Abstract:With the maturing of AI and multiagent systems research, we have a tremendous opportunity to direct these advances towards addressing complex societal problems. We focus on the problems of public health, wildlife conservation, and public safety, and present research advances in multiagent systems to address one key cross-cutting challenge: how to effectively deploy our limited intervention resources in these problem domains. We present our deployments from around the world as well as lessons learned that we hope are of use to researchers who are interested in AI for Social Impact. A key lesson in achieving social impact, we are automatically presented opportunities for advancing AI research; we focus in particular on research advances in influence maximization in social networks and computational game theory for conservation. In pushing this research agenda, we believe AI can indeed play an important role in fighting social injustice and improving society.
Bio:Milind Tambe is Gordon McKay Professor of Computer Science and Director of Center for Research in Computation and Society at Harvard University; concurrently, he is also Director "AI for Social Good" at Google Research India. He is recipient of the IJCAI John McCarthy Award, ACM/SIGAI Autonomous Agents Research Award, AAAI Robert S Engelmore Memorial Lecture award; he is also a fellow of AAAI and ACM. For his research in and pioneering real-world deployment of security games, Prof. Tambe has received the INFORMS Wagner prize, the Rist Prize of the Military Operations Research Society, the Christopher Columbus Fellowship Foundation Homeland security award, as well as Commendations from the US Coast Guard, LA Airport Police, and US Federal Air Marshals Service. Additionally, he and his team have received over a dozen influential paper and best paper awards at premier Artificial Intelligence Conferences. Prof. Tambe received his Ph.D. from the School of Computer Science at Carnegie Mellon University.
Bio:Milind Tambe is Gordon McKay Professor of Computer Science and Director of Center for Research in Computation and Society at Harvard University; concurrently, he is also Director "AI for Social Good" at Google Research India. He is recipient of the IJCAI John McCarthy Award, ACM/SIGAI Autonomous Agents Research Award, AAAI Robert S Engelmore Memorial Lecture award; he is also a fellow of AAAI and ACM. For his research in and pioneering real-world deployment of security games, Prof. Tambe has received the INFORMS Wagner prize, the Rist Prize of the Military Operations Research Society, the Christopher Columbus Fellowship Foundation Homeland security award, as well as Commendations from the US Coast Guard, LA Airport Police, and US Federal Air Marshals Service. Additionally, he and his team have received over a dozen influential paper and best paper awards at premier Artificial Intelligence Conferences. Prof. Tambe received his Ph.D. from the School of Computer Science at Carnegie Mellon University.
6.30 p.m. - 6.45 p.m
Concluding remarks