BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Halıcıoğlu Data Science Institute - UC San Diego - ECPv6.16.3//NONSGML v1.0//EN CALSCALE:GREGORIAN METHOD:PUBLISH X-WR-CALNAME:Halıcıoğlu Data Science Institute - UC San Diego X-ORIGINAL-URL:https://datascience.ucsd.edu X-WR-CALDESC:Events for Halıcıoğlu Data Science Institute - UC San Diego REFRESH-INTERVAL;VALUE=DURATION:PT1H X-Robots-Tag:noindex X-PUBLISHED-TTL:PT1H BEGIN:VTIMEZONE TZID:America/Los_Angeles BEGIN:DAYLIGHT TZOFFSETFROM:-0800 TZOFFSETTO:-0700 TZNAME:PDT DTSTART:20200308T100000 END:DAYLIGHT BEGIN:STANDARD TZOFFSETFROM:-0700 TZOFFSETTO:-0800 TZNAME:PST DTSTART:20201101T090000 END:STANDARD BEGIN:DAYLIGHT TZOFFSETFROM:-0800 TZOFFSETTO:-0700 TZNAME:PDT DTSTART:20210314T100000 END:DAYLIGHT BEGIN:STANDARD TZOFFSETFROM:-0700 TZOFFSETTO:-0800 TZNAME:PST DTSTART:20211107T090000 END:STANDARD BEGIN:DAYLIGHT TZOFFSETFROM:-0800 TZOFFSETTO:-0700 TZNAME:PDT DTSTART:20220313T100000 END:DAYLIGHT BEGIN:STANDARD TZOFFSETFROM:-0700 TZOFFSETTO:-0800 TZNAME:PST DTSTART:20221106T090000 END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210512T080000 DTEND;TZID=America/Los_Angeles:20210512T130000 DTSTAMP:20260607T212242 CREATED:20210505T211324Z LAST-MODIFIED:20210505T211324Z UID:10000323-1620806400-1620824400@datascience.ucsd.edu SUMMARY:Data Science Day DESCRIPTION: URL:https://bit.ly/3eZYr1s#new_tab CATEGORIES:Seminar,Webinar,Workshops ATTACH;FMTTYPE=: END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210506T130000 DTEND;TZID=America/Los_Angeles:20210506T140000 DTSTAMP:20260607T212242 CREATED:20210504T195111Z LAST-MODIFIED:20210504T195111Z UID:10000321-1620306000-1620309600@datascience.ucsd.edu SUMMARY:HDSI Seminar Series: Unavoidable Tensions in Explaining Algorithmic Decisions DESCRIPTION:Title: Unavoidable Tensions in Explaining Algorithmic Decisions \nAbstract: Recent developments in methods for explaining the decisions of machine learning models have been widely embraced for their ability to provide transparency and accountability without limiting model complexity or compelling model disclosure. Yet applying these methods is far from straightforward and they rarely prove a cure all. This talk identifies a number of unavoidable tensions that decision makers must navigate as they seek to employ these methods—and the deeply subjective judgment that must go into these considerations. \nPapers: This presentation is based–and builds–on two papers\, which are joint work with Andrew Selbst and Manish Raghavan: \n\nBarocas\, Solon\, Andrew D. Selbst\, and Manish Raghavan. “The hidden assumptions behind counterfactual explanations and principal reasons.” In Proceedings of the 2020 Conference on Fairness\, Accountability\, and Transparency\, pp. 80-89. 2020. https://dl.acm.org/doi/abs/10.1145/3351095.3372830\nSelbst\, Andrew D.\, and Solon Barocas. “The Intuitive Appeal of Explainable Machines.” Fordham Law Review 87\, no. 3 (2018): 1085. https://ir.lawnet.fordham.edu/cgi/viewcontent.cgi?article=5569&context=flr\n\nBio:  Solon Barocas is a Principal Researcher in the New York City lab of Microsoft Research\, an Adjunct Assistant Professor in the Department of Information Science at Cornell\, and Faculty Associate at the Berkman Klein Center for Internet & Society at Harvard. His research explores ethical and policy issues in artificial intelligence\, particularly fairness in machine learning\, methods for bringing accountability to automated decision-making\, and the privacy implications of inference. Solon co-founded the ACM conference on Fairness\, Accountability\, and Transparency (FAccT). URL:https://datascience.ucsd.edu/event/hdsi-seminar-series-unavoidable-tensions-in-explaining-algorithmic-decisions/ CATEGORIES:Guest Lecture,HDSI Event,Seminar ATTACH;FMTTYPE=: END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210503T170000 DTEND;TZID=America/Los_Angeles:20210503T180000 DTSTAMP:20260607T212242 CREATED:20210423T214856Z LAST-MODIFIED:20210423T214856Z UID:10000176-1620061200-1620064800@datascience.ucsd.edu SUMMARY:Seminar: Machine Learning for Spatio-Temporal Problems DESCRIPTION:Or Cohen\, Lyft \nMachine Learning for Spatio-Temporal Problems \nMay 3\, 2021\, 5pm: \nJoin Zoom Meeting \nhttps://ucsd.zoom.us/j/95036249217?pwd=a2kwSFg2ZXMvc21yMHZLRVR2c0pqUT09  \n  \nAbstract: \nMachine learning is used widely at Lyft. A few examples include predicting where and when ride requests will happen\, predicting travel time between two locations or predicting the probability for a passenger to cancel his/her ride. Such applications of machine learning are one of the key factors that differentiate Lyft from traditional taxi companies. In this talk\, I will present a few of these use-cases in detail. I will describe the unique challenges that come when applying machine learning for such spatio-temporal problems\, in which the main features are location (e.g. where the passenger is) and time (e.g. when the ride is requested). I will describe the different techniques for discretizing these variables\, which models work best for such problems and what challenges still remain. \n  \nBio: \nOr Cohen is Staff Data Scientist at Lyft focusing on ETA (Expected Time of Arrival). He has a PhD in statistical physics. \nhttps://www.deeplearning.ai/blog/working-ai-at-the-office-with-research-scientist-or-cohen/ \n  URL:https://datascience.ucsd.edu/event/seminar-machine-learning-for-spatio-temporal-problems/ CATEGORIES:Seminar ATTACH;FMTTYPE=: END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210503T130000 DTEND;TZID=America/Los_Angeles:20210503T140000 DTSTAMP:20260607T212242 CREATED:20210428T220805Z LAST-MODIFIED:20210428T220805Z UID:10000178-1620046800-1620050400@datascience.ucsd.edu SUMMARY:HDSI Seminar Series: Causal Effect Inference: A Machine Learning Approach by Mihaela van der Schaar DESCRIPTION:Title: Causal Effect Inference: A Machine Learning Approach \nAbstract \nA major challenge in the domain of healthcare is ascertaining whether a given treatment influences or determines an outcome—for instance\, whether there is a survival benefit to prescribing a certain medication. Current treatment guidelines have been developed with the “average” patient in mind\, but the reality is that treatments result in different effects and outcomes from one individual to another. \nUsing AI and machine learning\, we can endeavor to understand the effect of specific treatments on specific patients at specific times\, given their unique characteristics. This is what we call causal effect inference\, or individualized treatment effect inference. This is far from a straightforward undertaking\, however. In this seminar\, I will offer an introduction to individualized treatment effect inference for healthcare. I will explain the importance of this research area\, while also highlighting some key challenges\, formalisms\, methodologies\, and applications. \nBio \n \nMihaela van der Schaar is the John Humphrey Plummer Professor of Machine Learning\, Artificial Intelligence and Medicine at the University of Cambridge\, a Fellow at The Alan Turing Institute in London\, and a Chancellor’s Professor at UCLA. \nMihaela was elected IEEE Fellow in 2009. She has received numerous \nawards\, including the Oon Prize on Preventative Medicine from the University of Cambridge (2018)\, a National Science Foundation CAREER Award (2004)\, 3 IBM Faculty Awards\, the IBM Exploratory Stream Analytics Innovation Award\, the Philips Make a Difference Award and several best paper awards\, including the IEEE Darlington Award. \nMihaela’s work has also led to 35 USA patents (many widely cited and adopted in standards) and 45+ contributions to international standards for which she received 3 International ISO (International Organization for Standardization) Awards. \nIn 2019\, she was identified by National Endowment for Science\, Technology and the Arts as the most-cited female AI researcher in the UK. She was also elected as a 2019 “Star in Computer Networking and Communications” by N²Women. Her research expertise spans signal and image processing\, communication networks\, network science\, multimedia\, game theory\, distributed systems\, machine learning and AI. \nMihaela’s research focus is on machine learning\, AI and operations research for healthcare and medicine. \nIn addition to leading the van der Schaar Lab\, Mihaela is founder and director of the Cambridge Centre for AI in Medicine (CCAIM). URL:https://youtu.be/yhUItwTP02o#new_tab CATEGORIES:Guest Lecture,HDSI Event,Seminar END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210430T100000 DTEND;TZID=America/Los_Angeles:20210430T110000 DTSTAMP:20260607T212242 CREATED:20210219T230141Z LAST-MODIFIED:20210219T230141Z UID:10000314-1619776800-1619780400@datascience.ucsd.edu SUMMARY:Seminar: Bioinspired AI for Hierarchical Material Design: Modeling\, Design\, and Manufacturing DESCRIPTION:Zoom ID: https://cuboulder.zoom.us/j/2251625831\nPassword: i-aim \nMarkus Buehler\, MIT\n“Bioinspired AI for Hierarchical Material Design: Modeling\, Design\, and Manufacturing” \nAbstract:\nTBA \nAbout the Speaker:\nTBA  URL:https://datascience.ucsd.edu/event/seminar-manifold-learning-for-free-energy-surface-exploration-2021-04-30/ CATEGORIES:Seminar ATTACH;FMTTYPE=: END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210429T130000 DTEND;TZID=America/Los_Angeles:20210429T140000 DTSTAMP:20260607T212242 CREATED:20210427T184915Z LAST-MODIFIED:20210427T184915Z UID:10000177-1619701200-1619704800@datascience.ucsd.edu SUMMARY:HDSI Seminar Series: A Modern Take on Huber Regression by Po Ling Loh DESCRIPTION:Title: A modern take on Huber regression \nAbstract: \nIn the first part of the talk\, we discuss the use of a penalized Huber M-estimator for high-dimensional linear regression. We explain how a fairly straightforward analysis yields high-probability error bounds that hold even when the additive errors are heavy-tailed. However\, the parameter governing the shape of the Huber loss must be chosen in relation to the scale of the error distribution. We discuss how to use an adaptive technique\, based on Lepski’s method\, to overcome the difficulties traditionally faced by applying Huber M-estimation in a context where both location and scale are unknown. \nIn the second part of the talk\, we turn to a more complicated setting where both the covariates and responses may be heavy-tailed and/or adversarially contaminated. We show how to modify the Huber regression estimator by first applying an appropriate “filtering” procedure to the data based on the covariates. We prove that in low-dimensional settings\, this filtered Huber regression estimator achieves near-optimal error rates. We further show that the commonly used least trimmed squares and least absolute deviation estimators may similarly be made robust to contaminated covariates via the same covariate filtering step. This is based on joint work with Ankit Pensia and Varun Jog. \nBio: \nPo-Ling Loh received her PhD in Statistics from UC Berkeley in 2014. From 2014-2016\, she was an Assistant Professor of Statistics at the University of Pennsylvania. From 2016-2018\, she was an Assistant Professor of Electrical & Computer Engineering at UW-Madison\, and from 2019-2020\, she was an Associate Professor of Statistics at UW-Madison and a Visiting Associate Professor of Statistics at Columbia University. She began a position as a Lecturer in the Department of Pure Mathematics and Mathematical Statistics at the University of Cambridge in January 2021. Po-Ling’s current research interests include high-dimensional statistics\, robustness\, and differential privacy. She is a recipient of an NSF CAREER Award\, an ARO Young Investigator Award\, the IMS Tweedie and Bernoulli Society New Researcher Awards\, and a Hertz Fellowship. URL:https://youtu.be/5f-5psF6XbA#new_tab CATEGORIES:Guest Lecture,HDSI Event,Seminar END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210421T173000 DTEND;TZID=America/Los_Angeles:20210421T190000 DTSTAMP:20260607T212242 CREATED:20210408T232049Z LAST-MODIFIED:20210408T232049Z UID:10000173-1619026200-1619031600@datascience.ucsd.edu SUMMARY:Data Science Insights: Take a Hack at COVID! DESCRIPTION: URL:https://www.meetup.com/San-Diego-Machine-Learning/events/277120924/#new_tab CATEGORIES:Seminar ATTACH;FMTTYPE=: END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210416T100000 DTEND;TZID=America/Los_Angeles:20210416T110000 DTSTAMP:20260607T212242 CREATED:20210219T230141Z LAST-MODIFIED:20210219T230141Z UID:10000313-1618567200-1618570800@datascience.ucsd.edu SUMMARY:Seminar: Extracting the Structural\, Mechanical\, and Transport Properties of Nanostructured Materials DESCRIPTION:Zoom ID: https://cuboulder.zoom.us/j/2251625831\nPassword: i-aim \nMargaret Murnane\, CU Boulder\n“Extracting the Structural\, Mechanical\, and Transport Properties of Nanostructured Materials” \nAbstract:\nTBA \nAbout the Speaker:\nTBA URL:https://datascience.ucsd.edu/event/seminar-manifold-learning-for-free-energy-surface-exploration-2021-04-16/ CATEGORIES:Seminar ATTACH;FMTTYPE=: END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210415T130000 DTEND;TZID=America/Los_Angeles:20210415T140000 DTSTAMP:20260607T212242 CREATED:20210413T001611Z LAST-MODIFIED:20260421T193637Z UID:10000174-1618491600-1618495200@datascience.ucsd.edu SUMMARY:Distinguished Lecture Series: Thoughts and Efforts on AI Meeting Production DESCRIPTION:[vc_row][vc_column][vc_column_text css=””] \nThoughts and Efforts on AI Meeting Production\nAbstract: \nMachine Learning systems for complex tasks – such as controlling industrial manufacturing processes in real-time; or writing medical imaging case reports – are becoming increasingly sophisticated and consist of a large number of data\, model\, algorithm\, and system elements and modules. Traditional benchmark/leaderboard-driven bespoke approaches in the ML community are not suited to meet highly demanding industrial standards beyond algorithmic performance\, such as safety\, energy efficiency\, and scalability typically expected in production systems in the industry. In this talk\, I discuss some technical issues toward production- and industrial-AI from the following aspects: theoretical foundation for trustworthy and panoramic learning with all experiences; compositional strategies for building Pan-ML programs from Lego-like blocks; optimization methods for auto-tunning and federated learning; and systems framework for scaling up and scaling out ML productions. I will provide a few examples of our effects to address each of these challenges in the form of first principle formula\, new algorithms\, and software toolkits. \nBio: \nEric P. Xing is the President of the Mohamed bin Zayed University of Artificial Intelligence\, a Professor of Computer Science at Carnegie Mellon University\, and the Founder and Chairman of Petuum Inc.\, a 2018 World Economic Forum Technology Pioneer company that builds standardized artificial intelligence development platform and operating system for broad and general industrial AI applications. He completed his PhD in Computer Science at UC Berkeley. His main research interests are the development of machine learning and statistical methodology; and composable\, automatic\, and scalable computational systems\, for solving problems involving automated learning\, reasoning\, and decision-making in artificial\, biological\, and social systems. Prof Xing is a board member of the International Machine Learning Society; he has served as the Program Chair (2014) and General Chair (2019) of the International Conference of Machine Learning (ICML).[/vc_column_text][/vc_column][/vc_row] URL:https://datascience.ucsd.edu/event/distinguished-lecture-series-thoughts-and-efforts-on-ai-meeting-production/ END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210408T130000 DTEND;TZID=America/Los_Angeles:20210408T140000 DTSTAMP:20260607T212242 CREATED:20210331T191442Z LAST-MODIFIED:20210331T191442Z UID:10000172-1617886800-1617890400@datascience.ucsd.edu SUMMARY:HDSI Seminar Series: Deep Learning for Market Design: Fairness\, Robustness\, and Expressiveness by John Dickerson DESCRIPTION:Title\nDeep Learning for Market Design: Fairness\, Robustness\, and Expressiveness \nJohn Dickerson\, Assistant Professor of Computer Science\, University of Maryland; Chief Scientist\, Arthur AI \nAbstractThe design of revenue-maximizing auctions with strong incentive guarantees is a core concern of economic theory. Computational auctions enable online advertising\, sourcing\, spectrum allocation\, and myriad financial markets. Analytic progress in this space is notoriously difficult; since Myerson’s 1981 work characterizing single-item optimal auctions\, there has been limited progress outside of restricted settings. A recent paper by Dütting et al. circumvents analytic difficulties by applying deep learning techniques to\, instead\, approximate optimal auctions. Their RegretNet architecture can represent auctions with arbitrary numbers of items and participants; it is trained to be empirically strategyproof\, but the property is never exactly verified leaving potential loopholes for market participants to exploit. In parallel\, new research from Ilvento et al. and other groups has developed notions of fairness in the context of auction design. Inspired by these advances\, in this talk\, we discuss extensions of these techniques for approximating auctions using deep learning to address concerns of* fairness while maintaining high revenue and strong incentive guarantees;\n* certified robustness\, that is\, verification of claimed strategyproofness of deep learned auctions; and\n* expressiveness via different demand functions and other constraints.To enable that last point\, we propose a new architecture to learn incentive compatible\, revenue-maximizing auctions from sampled valuations\, which uses the Sinkhorn algorithm to perform a differentiable bipartite matching. Our new framework allows the network to learn strategyproof revenue-maximizing mechanisms in settings not learnable by the previous RegretNet architecture. \nThis talk covers hot-off-the-presses work led by PhD students Michael Curry\, Ping-yeh Chiang\, and Samuel Dooley\, and undergraduate students Kevin Kuo\, Uro Lyi\, Anthony Ostuni\, and Elizabeth Horishny. Papers have appeared at NeurIPS-20 or are currently under review; please check arXiv or get in touch for drafts. \nBio\nJohn P Dickerson is an Assistant Professor of Computer Science at the University of Maryland as well as Chief Scientist of Arthur AI\, an enterprise-focused AI/ML model monitoring firm. He is a recipient of awards such as the NSF CAREER Award\, IEEE Intelligent Systems AI’s 10 to Watch\, Google Faculty Research Award\, Google Research Scholar Award\, and paper awards and nominations at venues such as AAAI. His research centers on solving practical economic problems using techniques from computer science\, stochastic optimization\, and machine learning. He has worked extensively on theoretical and empirical approaches to organ exchange where his work has set policy at the UNOS nationwide kidney exchange; worldwide blood donation markets with Facebook; game-theoretic approaches to counter-terrorism and negotiation\, where his models have been deployed; and market design problems in industry (e.g.\, online advertising) through various startups. Dickerson received his PhD in computer science from Carnegie Mellon University. URL:https://youtu.be/nWmI2bs_mcs#new_tab CATEGORIES:Colloquium,HDSI Event,Seminar END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210402T100000 DTEND;TZID=America/Los_Angeles:20210402T110000 DTSTAMP:20260607T212242 CREATED:20210219T230141Z LAST-MODIFIED:20210219T230141Z UID:10000319-1617357600-1617361200@datascience.ucsd.edu SUMMARY:Seminar: Manifold Learning for Free Energy Surface Exploration DESCRIPTION:Zoom ID: https://cuboulder.zoom.us/j/2251625831\nPassword: i-aim \nYannis Kevrekidis\, Johns Hopkins University\n“Manifold Learning for Free Energy Surface Exploration” \nAbstract:\nTBA \nAbout the Speaker:\nTBA URL:https://datascience.ucsd.edu/event/seminar-manifold-learning-for-free-energy-surface-exploration/2021-04-02/ CATEGORIES:Seminar ATTACH;FMTTYPE=: END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210401T130000 DTEND;TZID=America/Los_Angeles:20210401T140000 DTSTAMP:20260607T212242 CREATED:20210331T183837Z LAST-MODIFIED:20210331T183837Z UID:10000171-1617282000-1617285600@datascience.ucsd.edu SUMMARY:HDSI Seminar Series: On the Implicit Bias of Stochastic Gradient Descent with Moderate Learning Rate by Quanquan Gu DESCRIPTION:Title: On the Implicit Bias of Stochastic Gradient Descent with Moderate Learning Rate \nHDSI Seminar Series \nQuanquan Gu\, Assistant Professor of Computer Science at UCLA \nAbstract: Understanding the algorithmic bias of stochastic gradient descent (SGD) is one of the key challenges in modern machine learning and deep learning theory. Most of the existing works\, however\, focus on very small or even infinitesimal learning rate regimes and fail to cover practical scenarios where the learning rate is moderate and annealing. In this talk\, I will introduce our attempt to characterize the particular regularization effect of SGD in the moderate learning rate regime by studying its behavior for optimizing an overparameterized linear regression problem. In this case\, SGD and GD are known to converge to the unique minimum-norm solution; however\, with the moderate and annealing learning rate\, we show that they exhibit different directional bias: SGD converges along the large eigenvalue directions of the data matrix\, while GD goes after the small eigenvalue directions. Furthermore\, we show that such directional bias does matter when early stopping is adopted\, where the SGD output is nearly optimal but the GD output is suboptimal. Our analysis explains several folk arts in practice used for SGD hyperparameter tuning\, such as (1) linearly scaling the initial learning rate with batch size; and (2) overrunning SGD with a high learning rate even when the loss stops decreasing. \nThis talk is based on joint work with Jingfeng Wu\, Difan Zou\, and Vladimir Braverman. \nBio: Quanquan Gu is an Assistant Professor of Computer Science at UCLA. His current research is in the area of artificial intelligence and machine learning\, with a focus on developing and analyzing nonconvex optimization algorithms for machine learning and building the theoretical foundations of deep learning. He received his Ph.D. degree in Computer Science from the University of Illinois at Urbana-Champaign in 2014. He is a recipient of the Yahoo! Academic Career Enhancement Award\, NSF CAREER Award\, Simons Berkeley Research Fellowship\, Adobe Data Science Research Award\, Salesforce Deep Learning Research Award\, and AWS Machine Learning Research Award. URL:https://youtu.be/qLGvn4RJwl4#new_tab CATEGORIES:Colloquium,HDSI Event,Seminar ATTACH;FMTTYPE=: END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210324T173000 DTEND;TZID=America/Los_Angeles:20210324T190000 DTSTAMP:20260607T212242 CREATED:20210319T224647Z LAST-MODIFIED:20210319T224647Z UID:10000169-1616607000-1616612400@datascience.ucsd.edu SUMMARY:Seminar: AutoNet: Automated Network Construction from Massive Text Corpora DESCRIPTION: URL:https://www.meetup.com/San-Diego-Machine-Learning/events/276913260/#new_tab CATEGORIES:Seminar ATTACH;FMTTYPE=: END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210319T100000 DTEND;TZID=America/Los_Angeles:20210319T110000 DTSTAMP:20260607T212242 CREATED:20210216T234644Z LAST-MODIFIED:20210216T234644Z UID:10000310-1616148000-1616151600@datascience.ucsd.edu SUMMARY:Seminar: Cerebro: A Layered Data Platform for Scalable Deep Learning DESCRIPTION:ZoomID: https://cuboulder.zoom.us/j/96007553384 \nPassword:i-aim \n\nArun Kumar\, UCSD\n“Cerebro: A Layered Data Platform for Scalable Deep Learning” \nAbstract:\nDeep learning (DL) is gaining popularity across many domains thanks to tools such as TensorFlow and easier access to GPUs. But building large-scale DL applications is still too resource-intensive and painful for all but the big tech firms. A key reason for this pain is the expensive model selection process needed to get DL to work well. Existing DL systems treat this process as an afterthought\, leading to massive resource wastage and a usability mess. To tackle these issues\, we present our vision of a first-of-its-kind data platform for scalable DL\, Cerebro\, inspired by lessons from the database world (https://adalabucsd.github.io/cerebro.html ). We elevate the DL model selection process with higher-level APIs already inherent in practice and devise a series of novel multi-query optimization techniques to substantially raise resource efficiency. In turn\, this can help improve scalability\, reduce resource costs and energy usage\, and improve DL user productivity. This talk will present our system design rationales and architecture\, our recent research and empirical results\, a discussion on how Cerebro is being applied to use cases in public health and enterprises\, and our future research plans. \nOverview paper (9pg): http://cidrdb.org/cidr2021/papers/cidr2021_paper25.pdf; Short video (10min): https://www.youtube.com/watch?v=8QfMvdlmdic \nAbout the Speaker:\nArun Kumar is an Assistant Professor in the Department of Computer Science and Engineering and the Halicioglu Data Science Institute at the University of California\, San Diego (https://adalabucsd.github.io). He is a member of the Database Lab and Center for Networked Systems and an affiliate member of the AI Group. His primary research interests are in data management and systems for machine learning/artificial intelligence-based data analytics. Systems and ideas based on his research have been released as part of the Apache MADlib open-source library\, shipped as part of products from Cloudera\, IBM\, Oracle\, and Pivotal\, and used internally by Facebook\, Google\, LogicBlox\, Microsoft\, and other companies. He is a recipient of two SIGMOD research paper awards\, a SIGMOD Research Highlight Award\, three distinguished reviewer awards from SIGMOD/VLDB\, the PhD dissertation award from UW-Madison CS\, an NSF CAREER Award\, a Hellman Fellowship\, a UCSD oSTEM Faculty of the Year Award\, and research award gifts from Google\, Oracle\, and VMware. URL:https://datascience.ucsd.edu/event/seminar-cerebro-a-layered-data-platform-for-scalable-deep-learning/ ATTACH;FMTTYPE=: END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210310T170000 DTEND;TZID=America/Los_Angeles:20210310T180000 DTSTAMP:20260607T212242 CREATED:20210126T002015Z LAST-MODIFIED:20230929T214012Z UID:10000300-1615395600-1615399200@datascience.ucsd.edu SUMMARY:STEMtorship Network: Lessons Learned: STEM Graduate Panel DESCRIPTION: URL:https://datascience.ucsd.edu/event/stemtorship-network-challenging-imposter-syndrome-2021-03-10/ CATEGORIES:Workshops END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210305T100000 DTEND;TZID=America/Los_Angeles:20210305T170000 DTSTAMP:20260607T212242 CREATED:20210224T180411Z LAST-MODIFIED:20210224T180411Z UID:10000318-1614938400-1614963600@datascience.ucsd.edu SUMMARY:Webinar: NSF SCALE MoDL (Mathematical and Scientific Foundations of Deep Learning) DESCRIPTION:Dear Colleagues\, \nThe Stimulating Collaborative Advances Leveraging Expertise in the Mathematical and Scientific Foundations of Deep Learning (SCALE MoDL) Management Team representing several NSF Divisions\, including MPS/DMS\, will be hosting a webinar scheduled for March 5\, 2021 starting at 1:00 PM Eastern Time. This is following up on the message we sent about last week regarding the NSF program solicitation NSF-21-561. Any members of the community who are interested in this funding opportunity are invited to participate in the webinar. \nThe webinar will have an open question and answer period for your questions submitted anonymously through the Zoom webinar Q&A feature. Participants may also submit questions in advance through the registration form or by sending an email to modl@nsf.gov. \nPlease use the registration link below to set up your participation. You are welcome to share this message with colleagues who might be interested. \n+++++++++++++++++++++++ \nParticipants should register (and may do so in advance) at the web page: \nhttps://nsf.zoomgov.com/webinar/register/WN_N9X5842dQ3-B6CkNvYXMwA \nAfter registering\, you will receive a confirmation email containing information about joining the webinar. \n+++++++++++++++++++++++ \nSCALE MoDL proposals are due by 5 p.m. submitter’s local time on May 12\, 2021. URL:https://datascience.ucsd.edu/event/webinar-nsf-scale-modl-mathematical-and-scientific-foundations-of-deep-learning/ CATEGORIES:Webinar ATTACH;FMTTYPE=: END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210301T100000 DTEND;TZID=America/Los_Angeles:20210301T170000 DTSTAMP:20260607T212242 CREATED:20210224T190853Z LAST-MODIFIED:20210224T190853Z UID:10000320-1614592800-1614618000@datascience.ucsd.edu SUMMARY:HDSI 3rd Anniversary Symposium DESCRIPTION: URL:https://datascience.ucsd.edu/news-and-events/events/3rdanniversary/ CATEGORIES:HDSI Event END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210225T120000 DTEND;TZID=America/Los_Angeles:20210225T130000 DTSTAMP:20260607T212242 CREATED:20210216T231037Z LAST-MODIFIED:20210216T231037Z UID:10000307-1614254400-1614258000@datascience.ucsd.edu SUMMARY:Seminar: A Deep Look into COVID and a New World of Innovation DESCRIPTION:The next event in our public Deep Look series is coming up on Feb. 25 at noon with A Deep Look into COVID and a New World of Innovation. Chancellor Khosla will be joining a panel highlighting groundbreaking initiatives across the UC San Diego campus and research labs. Please help us spread the word — all are invited!… In addition to the chancellor\, the event features Associate Professor Omar Akbari\, Professor Rommie Amaro and Senior Diagnostic Adviser Dan Kolk… \nAlong with global health and economic devastation\, the COVID-19 pandemic has forged an unprecedented path to research and education innovation. Bold scientific advances and cooperation led to a novel vaccine developed in record time\, groundbreaking tools for detecting viruses and a pioneering vision for safely educating students. Join us on Feb. 25 as we discuss the trailblazing insights and innovations that led to the broad success of UC San Diego’s Return to Learn program with Chancellor Khosla\, along with scientists leading groundbreaking innovations related to detecting and analyzing SARS-CoV-2\, as well as the future of at-home diagnostic testing in response to COVID-19. \nRegistration: https://covid-innovation.eventbrite.com URL:https://datascience.ucsd.edu/event/seminar-a-deep-look-into-covid-and-a-new-world-of-innovation/ CATEGORIES:Seminar ATTACH;FMTTYPE=: END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210224T173000 DTEND;TZID=America/Los_Angeles:20210224T183000 DTSTAMP:20260607T212242 CREATED:20210126T004649Z LAST-MODIFIED:20210126T004649Z UID:10000308-1614187800-1614191400@datascience.ucsd.edu SUMMARY:#Geisel50 Event at the Library | A Conversation with Kevin Young DESCRIPTION: URL:https://library.ucsd.edu/news-events/events/kevin-young/#new_tab CATEGORIES:Guest Lecture END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210224T170000 DTEND;TZID=America/Los_Angeles:20210224T180000 DTSTAMP:20260607T212242 CREATED:20210126T002015Z LAST-MODIFIED:20230929T213933Z UID:10000299-1614186000-1614189600@datascience.ucsd.edu SUMMARY:STEMtorship Network: Addressing Microaggressions in STEM DESCRIPTION: URL:https://datascience.ucsd.edu/event/stemtorship-network-challenging-imposter-syndrome-2021-02-24/ CATEGORIES:Workshops END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210219T120000 DTEND;TZID=America/Los_Angeles:20210219T130000 DTSTAMP:20260607T212242 CREATED:20210126T003804Z LAST-MODIFIED:20210126T003804Z UID:10000306-1613736000-1613739600@datascience.ucsd.edu SUMMARY:Explore Deloitte Series DESCRIPTION:You are invited to Explore Deloitte\, a series of educational sessions taking place this Winter to help students learn more about professional services and explore career opportunities at Deloitte. Each session will be 45 minutes and will provide an overview of our businesses\, professional development\, and culture. Following these sessions there will be opportunities to meet 1:1 with a Deloitte recruiter and learn more about the recruitment process. \nWelcome to the Explore Deloitte Series\nExplore Deloitte is a series of educational sessions taking place this spring semester to help you learn more about professional services and explore career opportunities at Deloitte. Sign up for Explore Deloitte sessions and continue your exploration with Deloitte by engaging with our team throughout the semester. Come Explore Deloitte! \n\n\n\n\n\n\n\n\nGet started now\nStep 1: Register for each event by clicking on individual event tiles below \nStep 2: Take Explore your Fit \nStep 3: Explore on your own URL:https://usrecruiting.deloitte.com/exploredeloitte#new_tab CATEGORIES:Webinar,Workshops ATTACH;FMTTYPE=: END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210219T100000 DTEND;TZID=America/Los_Angeles:20210219T110000 DTSTAMP:20260607T212242 CREATED:20210219T164932Z LAST-MODIFIED:20210219T164932Z UID:10000312-1613728800-1613732400@datascience.ucsd.edu SUMMARY:Seminar: Data-Driven Design of Microstructural Material Systems DESCRIPTION:Data Science for Materials Discovery \nData-Driven Design of Microstructural Material Systems \nProfessor Wei Chen\nNorthwestern University \nIn this talk\, we will introduce the state-of-the-art data science methods for designing heterogeneous nano- and microstructural materials and metamaterial systems such as polymer nanocomposites\, light-weight composite structures\, functional microelectronics\, and solar cells. Research developments in microstructure characterization and reconstruction\, machine learning of key microstructure features\, and data-driven Bayesian optimization using mixed-variable Gaussian process models\, will be described. \nFriday\, February 19\n10-11 am PDT / 11 am MST/ 1 pm EST \nhttps://cuboulder.zoom.us/j/96007553384\nPassword: i-aim \nAbstract:\nDesign of advanced material systems imposes challenges in integrating knowledge and representation from multiple disciplines and domains such as materials\, manufacturing\, structural mechanics\, and design optimization. Data-driven machine learning and computational design methods provide a seamless integration of predictive materials modeling\, manufacturing\, and design optimization to enable the accelerated design and deployment of advanced materials systems. In this talk\, we will introduce the state-of-the-art data science methods for designing heterogeneous nano- and microstructural materials and metamaterial systems such as polymer nanocomposites\, light-weight composite structures\, functional microelectronics\, and solar cells. Research developments in microstructure characterization and reconstruction\, machine learning ​of key microstructure features\, and data-driven Bayesian optimization using mixed-variable Gaussian process models\, will be introduced. Challenges and opportunities in designing engineered material systems will be discussed. \nAbout the Speaker:\nWei Chen is the Wilson-Cook Professor in Engineering Design and Chair of Department of Mechanical Engineering at Northwestern University. Directing the Integrated DEsign Automation Laboratory (IDEAL- http://ideal.mech.northwestern.edu/)\, her current research involves issues such as simulation-based design under uncertainty; model validation and uncertainty quantification; data science in design and advanced manufacturing; stochastic multiscale analysis and materials design; design of metamaterials; multidisciplinary design optimization; consumer choice modeling and decision-based design.  She is a member of the National Academy of Engineering (NAE) and currently serving as the Editor-in-chief of the ASME Journal of Mechanical Design and the President of the International Society of Structural and Multidisciplinary Design (ISSMO). Dr. Chen was the recipient of the 2015 ASME Design Automation Award and 1996 NSF Faculty Career Award.  She received her Ph.D. from the Georgia Institute of Technology in 1995. URL:https://datascience.ucsd.edu/event/seminar-data-driven-design-of-microstructural-material-systems/ CATEGORIES:Seminar ATTACH;FMTTYPE=: END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210218T113000 DTEND;TZID=America/Los_Angeles:20210218T123000 DTSTAMP:20260607T212242 CREATED:20210218T163536Z LAST-MODIFIED:20210218T163536Z UID:10000311-1613647800-1613651400@datascience.ucsd.edu SUMMARY:Seminar: A Theory of Optimization and Transition to Linearity in Deep Learning by Mikhail Belkin DESCRIPTION:We are delighted that Misha Belkin will give this week’s One World Mathematics of Information\, Data\, and Signals (1W-MINDS) Seminar on Thursday\, February 18\, at 2:30 pm EDT (11:30 am Pacific time).  As usual\, attendees can participate in the seminar using the following zoom link at that time: \nhttps://msu.zoom.us/j/96421373881 (the passcode is the first prime number > 100). \nProf. Belkin’s title and abstract are below\, and can also be found on the seminar website along with information about other upcoming talks\, videos of past talks\, and more at https://sites.google.com/view/minds-seminar/home \n==========================================================\nA Theory of Optimization and Transition to Linearity in  Deep Learning \nThe success of deep learning is due\, to a large extent\, to the remarkable effectiveness of gradient-based optimization methods applied to large neural networks. In this talk I will discuss some general mathematical principles allowing for efficient optimization in over-parameterized non-linear systems\, a setting that includes deep neural networks. Remarkably\, it seems that optimization of such systems is “easy”. In particular\, optimization problems corresponding to these systems are not convex\, even locally\, but instead satisfy locally the Polyak-Lojasiewicz (PL) condition allowing for efficient optimization by gradient descent or SGD. We connect the PL condition of these systems to the condition number associated to the tangent kernel and develop a non-linear theory parallel to classical analyses of over-parameterized linear equations. \nIn a related but conceptually separate development\, I will discuss a new perspective on the remarkable recently discovered phenomenon of transition to linearity (constancy of NTK) in certain classes of large neural networks. I will show how this transition to linearity results from the scaling of the Hessian with the size of the network. \nJoint work with Chaoyue Liu and Libin Zhu URL:https://datascience.ucsd.edu/event/a-theory-of-optimization-and-transition-to-linearity-in-deep-learning/ CATEGORIES:Seminar END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210212T120000 DTEND;TZID=America/Los_Angeles:20210212T130000 DTSTAMP:20260607T212242 CREATED:20210126T003804Z LAST-MODIFIED:20210126T003804Z UID:10000305-1613131200-1613134800@datascience.ucsd.edu SUMMARY:Explore Deloitte Series DESCRIPTION:You are invited to Explore Deloitte\, a series of educational sessions taking place this Winter to help students learn more about professional services and explore career opportunities at Deloitte. Each session will be 45 minutes and will provide an overview of our businesses\, professional development\, and culture. Following these sessions there will be opportunities to meet 1:1 with a Deloitte recruiter and learn more about the recruitment process. \nWelcome to the Explore Deloitte Series\nExplore Deloitte is a series of educational sessions taking place this spring semester to help you learn more about professional services and explore career opportunities at Deloitte. Sign up for Explore Deloitte sessions and continue your exploration with Deloitte by engaging with our team throughout the semester. Come Explore Deloitte! \n\n\n\n\n\n\n\n\nGet started now\nStep 1: Register for each event by clicking on individual event tiles below \nStep 2: Take Explore your Fit \nStep 3: Explore on your own URL:https://usrecruiting.deloitte.com/exploredeloitte#new_tab CATEGORIES:Webinar,Workshops ATTACH;FMTTYPE=: END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210210T130000 DTEND;TZID=America/Los_Angeles:20210210T140000 DTSTAMP:20260607T212242 CREATED:20210210T200512Z LAST-MODIFIED:20210210T200512Z UID:10000309-1612962000-1612965600@datascience.ucsd.edu SUMMARY:Data Science Student Lightning Talks DESCRIPTION:UC Love Data Week’s session of data science student lightning talks. Join us to hear more about the projects these talented students\, all of whom are from UC San Diego\, have been working on. \n  \nDate: (February 10) \nTime: 1-2pm \nZoom link: https://ucsd.zoom.us/j/92839827420 \n  \nLightning talks: \n  \n\n\n\nPresenter name\nProposed talk title\n\n\nGil Pasternak\nLikelihood – Reimagining Data Quality\n\n\nArunav Gupta\nQuantifying the legacy of redlining policies in the urban landscape of modern-day America\n\n\nAdvitya Gemawat\nHyperparameter Optimization for Deep Neural Networks with Apache MADlib\, an open source ML library for scalable in-database analytics\n\n\nTed Feng\nAnalysis of the NBA Bubble\n\n\nBailey Man\nAnalysis of the retreat of global Ice caps\n\n\nYifei Ning\, Zhaoyi Hou\nUSPTO Patent Application Prediction – Dealing with Extremely Unbalanced Text Data URL:https://datascience.ucsd.edu/event/data-science-student-lightning-talks/ CATEGORIES:Workshops ATTACH;FMTTYPE=: END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210130T090000 DTEND;TZID=America/Los_Angeles:20210130T100000 DTSTAMP:20260607T212242 CREATED:20210125T234037Z LAST-MODIFIED:20210125T234037Z UID:10000293-1611997200-1612000800@datascience.ucsd.edu SUMMARY:DSGO Connect DESCRIPTION:Speaker Jack Raifer: Data Scientist and Behavioral Economist working on developing empathic\, and pro human Machine Learning and Artificial Intelligence systems for People Analytics and Human Development.Can you go from barely knowing what Data Science is\, to becoming a Data Scientist? The answer is yes! Join Jack Raifer Baruch to learn tips to make your road to Data Science smoother. After the presentation\, there will be discussion roundtables and networking. Don’t miss the opportunity to build powerful connections\, learn about career opportunities\, and grow your professional network \nDSGO CONNECTOur unique format breaks the mold of your typical data science event\, giving you more chances to connect with and learn from the industry’s top experts. Get closer and grow faster than you ever could before at DSGO Connect. \nStart living and breathing data science all year round via free monthly events that capture the cutting-edge energy of the industry\, help you grow closer to the data science community\, and build the skills you need to succeed in the real world. \n\n Speed Networking Sessions \n Meet more of your industry peers\, build your network in less time\, and learn from their experience. \n\n Industry Expert Q&As \n Top industry leaders are ready to answer your pressing questions on anything from to career advice\, to model building and everything in between. \n\n Panel Discussions \n Gain new perspectives on several topics from experts in the field.\n \n Register here URL:https://datascience.ucsd.edu/event/dsgo-connect/ CATEGORIES:Webinar,Workshops END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210129T120000 DTEND;TZID=America/Los_Angeles:20210129T130000 DTSTAMP:20260607T212242 CREATED:20210126T003804Z LAST-MODIFIED:20210126T003804Z UID:10000304-1611921600-1611925200@datascience.ucsd.edu SUMMARY:Explore Deloitte Series DESCRIPTION:You are invited to Explore Deloitte\, a series of educational sessions taking place this Winter to help students learn more about professional services and explore career opportunities at Deloitte. Each session will be 45 minutes and will provide an overview of our businesses\, professional development\, and culture. Following these sessions there will be opportunities to meet 1:1 with a Deloitte recruiter and learn more about the recruitment process. \nWelcome to the Explore Deloitte Series\nExplore Deloitte is a series of educational sessions taking place this spring semester to help you learn more about professional services and explore career opportunities at Deloitte. Sign up for Explore Deloitte sessions and continue your exploration with Deloitte by engaging with our team throughout the semester. Come Explore Deloitte! \n\n\n\n\n\n\n\n\nGet started now\nStep 1: Register for each event by clicking on individual event tiles below \nStep 2: Take Explore your Fit \nStep 3: Explore on your own URL:https://usrecruiting.deloitte.com/exploredeloitte#new_tab CATEGORIES:Webinar,Workshops ATTACH;FMTTYPE=: END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210127T170000 DTEND;TZID=America/Los_Angeles:20210127T180000 DTSTAMP:20260607T212242 CREATED:20210126T002015Z LAST-MODIFIED:20240501T192654Z UID:10000298-1611766800-1611770400@datascience.ucsd.edu SUMMARY:STEMtorship Network: Challenging Imposter Syndrome DESCRIPTION: URL:https://datascience.ucsd.edu/event/stemtorship-network-challenging-imposter-syndrome/ CATEGORIES:Workshops END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210127T150000 DTEND;TZID=America/Los_Angeles:20210127T170000 DTSTAMP:20260607T212242 CREATED:20210125T232325Z LAST-MODIFIED:20210125T232325Z UID:10000292-1611759600-1611766800@datascience.ucsd.edu SUMMARY:Million Dollar Hoods: Mapping the Fiscal and Human Cost of Mass Incarceration in Los Angeles by Kelly Lytle Hernandez DESCRIPTION:Ethnic Studies 30th Anniversary Event\nAbstract: Los Angeles County operates the largest jail system in the United States\, which incarcerates more people than any other nation on Earth. This talk provides an introduction to the Million Dollar Hoods(MDH) project\, method\, and impact. Led by Professor Lytle\, MDH is a university-based\, community-driven research project that maps the fiscal and human cost of mass incarceration in Los Angeles. \nBio: Kelly Lytle Hernandez: An Ethnic Studies undergraduate alumnus\, Kelly Lytle Hernandez is a professor of History\, African American Studies\, and Urban Planning at UCLA where she holds The Thomas E. Lifka Endowed Chair in History. She is also the Director of the Ralph J Bunche Center for African American Studies at UCLA. One of the nation’s leading experts on race. immigration\, and mass incarceration\, she is the author of th eaward-winning books\, Migra! A History of the US Border Patrol and City of Inmates: Conquest. Rebellion\, and the Rise of Human Caging In Los Angeles. Currently. Professor Lytle Hernandez Is the Director and Principal Investigator for Million Dollar Hoods. a research project that maps the fiscal and human cost of mass Incarceration in Los Angeles. In 2019. she was named a MacArthur ‘Genius’ Fellow. \nDiscussants:\n\nMohamed Abuyame (Ethnic Studies PhD alumnus\, 2017): Mohamed Abumaye is Assistant Professor in the Department of Sociology at California State University\, San Marcos. His work centers on the intersections between military and police violence\, centering the role of Somali youth activists in exposing the relationship between U.S. militarism in Somalia and hyper-policing in City Heights\, San Diego. \nJesse Mills (Ethnic Studies PhD alumnus\, 2008): Jesse Mills is Associate Professor and former Chair of Ethnic Studies at the University of San Diego. He has conducted research and volunteered in the City Heights Somali community since 2002\, where he has served as ESL teacher\, youth mentor\, educational consultant\, organizational advisor\, and community member. URL:https://datascience.ucsd.edu/event/million-dollar-hoods-mapping-the-fiscal-and-human-cost-of-mass-incarceration-in-los-angeles-by-kelly-lytle-hernandez/ CATEGORIES:Seminar ATTACH;FMTTYPE=: END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20210122T140000 DTEND;TZID=America/Los_Angeles:20210122T153000 DTSTAMP:20260607T212242 CREATED:20210126T000256Z LAST-MODIFIED:20230929T214911Z UID:10000295-1611324000-1611329400@datascience.ucsd.edu SUMMARY:The Plug: UC Scholars/Genentech Scholars/Undergraduate Research Scholarships DESCRIPTION: URL:https://datascience.ucsd.edu/event/the-plug-virtual-workshops-hosted-by-the-writing-hub-and-black-resource-center/ CATEGORIES:Workshops END:VEVENT END:VCALENDAR