BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Halıcıoğlu Data Science Institute - UC San Diego - ECPv6.16.2//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:20230312T100000 END:DAYLIGHT BEGIN:STANDARD TZOFFSETFROM:-0700 TZOFFSETTO:-0800 TZNAME:PST DTSTART:20231105T090000 END:STANDARD BEGIN:DAYLIGHT TZOFFSETFROM:-0800 TZOFFSETTO:-0700 TZNAME:PDT DTSTART:20240310T100000 END:DAYLIGHT BEGIN:STANDARD TZOFFSETFROM:-0700 TZOFFSETTO:-0800 TZNAME:PST DTSTART:20241103T090000 END:STANDARD BEGIN:DAYLIGHT TZOFFSETFROM:-0800 TZOFFSETTO:-0700 TZNAME:PDT DTSTART:20250309T100000 END:DAYLIGHT BEGIN:STANDARD TZOFFSETFROM:-0700 TZOFFSETTO:-0800 TZNAME:PST DTSTART:20251102T090000 END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240405T123000 DTEND;TZID=America/Los_Angeles:20240405T140000 DTSTAMP:20260531T085249 CREATED:20240329T001842Z LAST-MODIFIED:20240329T001950Z UID:10000468-1712320200-1712325600@datascience.ucsd.edu SUMMARY:"Advancing NLP for Timely and Actionable Feedback in Healthcare Conversations" | Veronica Perez-Rosas DESCRIPTION:Abstract: “Effective communication is crucial in healthcare for ensuring successful clinical interactions\, as it affects how patients respond\, the decisions being made by both patients and clinicians\, and the outcomes of treatments. Recent developments in Natural Language Processing (NLP) aim to improve and support these interactions within clinical settings. In this talk\, I will discuss my research on offering timely and actionable evaluative feedback for mental healthcare interactions\, addressing a crucial bottleneck in effective mental healthcare delivery. I will specifically focus on computational approaches for building conversational systems to aid in psychotherapy training\, and present two NLP tasks to generate language-based feedback: (1) generating counselor responses following established counseling strategies\, and (2) offering alternative rewrites to counseling trainees’ responses to refine their counseling skills. I will conclude the talk by outlining future directions towards my long-term agenda of building computational approaches that understand\, model\, and predict health behaviors while also being human-centric and scalable” \nBio: “Veronica Perez-Rosas is an Assistant Research Scientist at the University of Michigan. She received her Ph.D. in Computer Science and Engineering from the University of North Texas in 2014\, and was a postdoctoral fellow at the University of Michigan until 2016. Her research interests include Natural Language Processing\, Machine Learning\, Affect Recognition\, and Multimodal Processing of Human Behavior. Her research focuses on developing computational methods to analyze\, recognize\, and predict human behaviors during social interactions. She has authored papers in leading conferences and journals in Natural Language Processing and Multimodal Processing\, has mentored numerous students in these research areas\, and has served as workshop chair or area chair for multiple international conferences in the field.” URL:https://datascience.ucsd.edu/event/advancing-nlp-for-timely-and-actionable-feedback-in-healthcare-conversations-veronica-perez-rosas/ LOCATION:Halıcıoğlu Data Science Institute (HDSI)\, Room 123\, 3234 Matthews Ln\, La Jolla\, CA\, 92093\, United States CATEGORIES:Seminar END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240404T080000 DTEND;TZID=America/Los_Angeles:20240405T120000 DTSTAMP:20260531T085249 CREATED:20240226T234322Z LAST-MODIFIED:20240313T192254Z UID:10000418-1712217600-1712318400@datascience.ucsd.edu SUMMARY:Causality Workshop DESCRIPTION: URL:https://www.eventbrite.com/e/ucsd-hdsi-causality-workshop-tickets-817326594847?aff=oddtdtcreator LOCATION:SDSC\, The Auditorium\, 9836 Hopkins Dr\, La Jolla\, San Diego\, CA\, United States CATEGORIES:Workshops ATTACH;FMTTYPE=image/png:https://datascience.ucsd.edu/wp-content/uploads/2024/01/HDSI_Causality_Wrkshp_Eventbrite.png END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240403T140000 DTEND;TZID=America/Los_Angeles:20240403T153000 DTSTAMP:20260531T085249 CREATED:20240326T221709Z LAST-MODIFIED:20240329T001345Z UID:10000462-1712152800-1712158200@datascience.ucsd.edu SUMMARY:"Contextualized learning for adaptive yet persistent AI in biomedicine" | Ben Lengerich DESCRIPTION:Abstract: “In biomedical data analysis\, an emerging trend focuses on contextualizing observations within biological and real-world processes. This approach facilitates high-resolution\, context-specific insights by integrating information across datasets\, but it is difficult to design systems which both share information and dynamically adapt to context. Toward this aim\, this presentation will examine “contextualized learning”\, a meta-learning paradigm which learns relationships between dataset context and statistical parameters. Using contextualized network inference as an illustrative example\, I will show how we can estimate context-specific graphical models\, offering insights such as personalized gene expression analysis for SOTA cancer subtyping. The talk will also discuss trends towards “contextualized understanding”\, bridging statistical and foundation models to standardize interpretability. The primary aim is to illustrate how contextualized learning and understanding contribute to creating learning systems that are both adaptive and persistent\, facilitating cross-context information sharing and detailed analysis.” \nBio: “Ben Lengerich is a Postdoctoral Associate and Alana Fellow at MIT’s Computer Science and Artificial Intelligence Lab (CSAIL) and the Broad Institute of MIT and Harvard\, where he is advised by Manolis Kellis. His research in machine learning and computational biology emphasizes the use of context-adaptive models to understand complex diseases and advance precision medicine. Through his work\, Ben aims to bridge the gap between data-driven insights and actionable medical interventions. He holds a PhD in Computer Science and MS in Machine Learning from Carnegie Mellon University\, where he was advised by Eric Xing. His work has been recognized with spotlight presentations at conferences including NeurIPS\, ISMB\, AMIA\, and SMFM\, financial support from the Alana Foundation\, selection as a “”Rising Star in Data Science” by the University of Chicago and UC San Diego\, and “”Next Generation in Biomedicine”” by the Broad Institute.” URL:https://datascience.ucsd.edu/event/special-seminar-ben-lengerich/ LOCATION:Computer Science & Engineering Building (CSE)\, Room 1202 CATEGORIES:Seminar ATTACH;FMTTYPE=image/png:https://datascience.ucsd.edu/wp-content/uploads/2024/01/HDSI-UCSD-Image_Dark-blue-e1710178042629.png END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240403T120000 DTEND;TZID=America/Los_Angeles:20240403T133000 DTSTAMP:20260531T085249 CREATED:20240327T215044Z LAST-MODIFIED:20240401T225533Z UID:10000466-1712145600-1712151000@datascience.ucsd.edu SUMMARY:MathWorks & HDSI AI Seminar | Esperanza Linares DESCRIPTION:HDSI! Come and join MathWorks Engineers for a technical seminar on AI (and lunch!) on Wednesday\, April 3! Come learn why data scientists should learn MATLAB – we will highlight the tools that will be serve your role as data scientists and data science students. You can also learn about our engineer’s journey\, roles available at MathWorks\, and the use of our tools in industry! \nMathworks UCSD Technical Seminar Series \nLow-Code AI in MATLAB \nLearn how you can apply AI in your field without extensive knowledge in programming. This hands-on session includes a quick recap on the fundamentals of AI and three exercises where you will learn how to classify human activities using MATLAB® interactive tools and apps: \n1. Accessing and preprocessing data acquired from a mobile device\n2. Applying clustering to the unlabeled data using the Cluster Data Live Editor Task\n3. Classifying the labeled data using two apps: Classification Learner app and the Deep Network Designer app \nAt the end of the seminar\, you will be able to design and train different machine learning and deep learning models without extensive programming knowledge. You will also learn how to automatically generate code from the interactive workflow. This will not only help you to reuse the models without manually going through all the steps but also to learn programming or advance your coding skills. \nAbout the Speaker: \nEsperanza Linares is a Senior Customer Success Engineer at MathWorks. She is part of a global team that partners with academic and research institutions worldwide\, focusing on student and research success. Before joining MathWorks\, she did her postdoctoral work in the pharmaceutical industry\, where she developed a discrete element method model to simulate the compaction of granular materials. She holds a BS in Mechanical Engineering from UNAM (Mexico) and a Ph.D. in Mechanical Engineering from Caltech. \nRegistration Link: https://forms.office.com/Pages/ResponsePage.aspx?id=ETrdmUhDaESb3eUHKx3B5tTIy0i-nn1KjKWuEYZzK09UNVNXNFM4NTA3Q045REVJWUNHNjcxUkZSTi4u \n*Lunch will be provided URL:https://datascience.ucsd.edu/event/mathworks-hdsi-ai-seminar-esperanza-linares/ LOCATION:Halıcıoğlu Data Science Institute (HDSI)\, Room 123\, 3234 Matthews Ln\, La Jolla\, CA\, 92093\, United States CATEGORIES:Seminar ATTACH;FMTTYPE=image/png:https://datascience.ucsd.edu/wp-content/uploads/2023/03/mathworks_logo.png END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240402T140000 DTEND;TZID=America/Los_Angeles:20240402T153000 DTSTAMP:20260531T085249 CREATED:20240313T191528Z LAST-MODIFIED:20240313T191528Z UID:10000459-1712066400-1712071800@datascience.ucsd.edu SUMMARY:Special Seminar | Xuhai Xu DESCRIPTION: URL:https://datascience.ucsd.edu/event/special-seminar-xuhai-xu/ LOCATION:Halıcıoğlu Data Science Institute (HDSI)\, Room 123\, 3234 Matthews Ln\, La Jolla\, CA\, 92093\, United States CATEGORIES:Seminar END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240401T140000 DTEND;TZID=America/Los_Angeles:20240401T153000 DTSTAMP:20260531T085249 CREATED:20240304T172031Z LAST-MODIFIED:20240329T000153Z UID:10000453-1711980000-1711985400@datascience.ucsd.edu SUMMARY:"Instance-Optimization: Rethinking Database Design for the Next 1000X" | Jialin Ding DESCRIPTION:Abstract: “Modern database systems aim to support a large class of different use cases while simultaneously achieving high performance. However\, as a result of their generality\, databases often achieve adequate performance for the average use case but do not achieve the best performance for any individual use case. In this talk\, I will describe my work on designing databases that use machine learning and optimization techniques to automatically achieve performance much closer to the optimal for each individual use case. In particular\, I will present my work on instance-optimized database storage layouts\, in which the co-design of data structures and optimization policies improves query performance in analytic databases by orders of magnitude. I will highlight how these instance-optimized data layouts address various challenges posed by real-world database workloads and how I implemented and deployed them in production within Amazon Redshift\, a widely-used commercial database system.” \nBio: “Jialin Ding is an Applied Scientist at AWS. Before that\, he received his PhD in computer science from MIT\, advised by Tim Kraska. He works broadly on applying machine learning and optimization techniques to improve data management systems\, with a focus on building databases that automatically self-optimize to achieve high performance for any specific application. His work has appeared in top conferences such as SIGMOD\, VLDB\, and CIDR\, and has been recognized by a Meta Research PhD Fellowship. To learn more about Jialin’s work\, please visit https://jialinding.github.io/.” URL:https://datascience.ucsd.edu/event/special-seminar-jialin-ding/ LOCATION:Halıcıoğlu Data Science Institute (HDSI)\, Room 123\, 3234 Matthews Ln\, La Jolla\, CA\, 92093\, United States CATEGORIES:Seminar ATTACH;FMTTYPE=image/png:https://datascience.ucsd.edu/wp-content/uploads/2024/01/HDSI-UCSD-Image_Dark-blue-e1710178042629.png END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240401T110000 DTEND;TZID=America/Los_Angeles:20240401T123000 DTSTAMP:20260531T085249 CREATED:20240328T234406Z LAST-MODIFIED:20240328T234737Z UID:10000467-1711969200-1711974600@datascience.ucsd.edu SUMMARY:How Do We Get There?: Toward Intelligent Behavior Intervention | Xuhai Xu DESCRIPTION:Abstract: As the intelligence of everyday smart devices continues to evolve\, they can already monitor basic health behaviors such as physical activities and heart rates. The vision of an intelligent behavior change intervention pipeline for health — combining behavior modeling & interaction design — seems to be within reach. How do we get there? \nIn this talk\, I will introduce a comprehensive intervention pipeline that bridges behavior science theory-driven designs and generalizable behavior models. I will also introduce my efforts on passive sensing datasets\, human-centered algorithms\, and a benchmark platform that drives the community toward more robust and deployable intervention systems for health and well-being. \nBio: Xuhai “Orson” Xu is a postdoc at MIT EECS. He received his PhD at the University of Washington. Specializing in human-computer interaction\, applied machine learning\, and health\, Xu develops intelligent behavior intervention systems to promote human health and well-being. His research covers two aspects — 1) building deployable human-centered behavior models and 2) designing interactive user experiences — to establish a complete system to improve end-users’ well-being. Moreover\, his research also goes beyond end-users and supports health experts by designing new human-AI collaboration paradigms in clinical settings. Xu has earned several awards\, including 9 Best Paper\, Best Paper Honorable Mention\, and Best Artifact awards. His research has been covered by media outlets such as the Washington Post and ACM News. He was recognized as the Outstanding Student Award Winner at UbiComp 2022\, the 2023 UW Distinguished Dissertation Award\, and the 2024 Innovation and Technology Award at the Western Association of Graduate Schools. \nZoom: https://ucsd.zoom.us/j/92792843021\nPassword: 741675 URL:https://datascience.ucsd.edu/event/how-do-we-get-there-toward-intelligent-behavior-intervention-xuhai-xu/ LOCATION:Computer Science & Engineering Building (CSE)\, Room 1242\, 3234 Matthews Ln\, La Jolla\, CA\, 92093\, United States CATEGORIES:Seminar ATTACH;FMTTYPE=image/png:https://datascience.ucsd.edu/wp-content/uploads/2024/01/HDSI-UCSD-Image_Dark-blue-e1710178042629.png END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240328T140000 DTEND;TZID=America/Los_Angeles:20240328T153000 DTSTAMP:20260531T085249 CREATED:20240326T031112Z LAST-MODIFIED:20240326T031112Z UID:10000464-1711634400-1711639800@datascience.ucsd.edu SUMMARY:The Emergence of Reproducibility and Generalizability in Diffusion Models | Qing Qu DESCRIPTION:Abstract: We reveal an intriguing and prevalent phenomenon of diffusion models which we term as “consistent model reproducibility”: given the same starting noise input and a deterministic sampler\, different diffusion models often yield remarkably similar outputs while they generate new samples. We demonstrate this phenomenon through comprehensive experiments and theoretical studies\, implying that different diffusion models consistently reach the same data distribution and scoring function regardless of frameworks\, model architectures\, or training procedures. More strikingly\, our further investigation implies that diffusion models are learning distinct distributions affected by the training data size and model capacity\, so that the model reproducibility manifests in two distinct training regimes with phase transition: (i) “memorization regime”\, where the diffusion model overfits to the training data distribution\, and (ii) “generalization regime”\, where the model learns the underlying data distribution and generate new samples with finite training data. Finally\, our results have strong practical implications regarding training efficiency\, model privacy\, and controllable generation of diffusion models\, and our work raises numerous intriguing theoretical questions for future investigation. \nSpeaker Bio: Qing Qu is an assistant professor in EECS department at the University of Michigan. Prior to that\, he was a Moore-Sloan data science fellow at Center for Data Science\, New York University\, from 2018 to 2020. He received his Ph.D from Columbia University in Electrical Engineering in Oct. 2018. He received his B.Eng. from Tsinghua University in Jul. 2011\, and a M.Sc.from the Johns Hopkins University in Dec. 2012\, both in Electrical and Computer Engineering. His research interest lies at the intersection of foundation of data science\, machine learning\, numerical optimization\, and signal/image processing\, with focus on developing efficient nonconvex methods and global optimality guarantees for solving representation learning and nonlinear inverse problems in engineering and imaging sciences. He is the recipient of Best Student Paper Award at SPARS’15\, and the recipient of Microsoft PhD Fellowship in machine learning in 2016\, and best paper awards in NeurIPS Diffusion Model Workshop in 2023. He received the NSF Career Award in 2022\, and Amazon Research Award (AWS AI) in 2023. He is the program chair of the new Conference on Parsimony & Learning. URL:https://datascience.ucsd.edu/event/the-emergence-of-reproducibility-and-generalizability-in-diffusion-models-qing-qu/ LOCATION:Halıcıoğlu Data Science Institute (HDSI)\, Room 123\, 3234 Matthews Ln\, La Jolla\, CA\, 92093\, United States CATEGORIES:Seminar ATTACH;FMTTYPE=image/png:https://datascience.ucsd.edu/wp-content/uploads/2024/01/HDSI-UCSD-Image_Dark-blue-e1710178042629.png END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240328T140000 DTEND;TZID=America/Los_Angeles:20240328T153000 DTSTAMP:20260531T085249 CREATED:20240304T171827Z LAST-MODIFIED:20240323T082150Z UID:10000452-1711634400-1711639800@datascience.ucsd.edu SUMMARY:The Emergence of Reproducibility and Generalizability in Diffusion Models | Qing Qu DESCRIPTION:Abstract: We reveal an intriguing and prevalent phenomenon of diffusion models which we term as “consistent model reproducibility”: given the same starting noise input and a deterministic sampler\, different diffusion models often yield remarkably similar outputs while they generate new samples. We demonstrate this phenomenon through comprehensive experiments and theoretical studies\, implying that different diffusion models consistently reach the same data distribution and scoring function regardless of frameworks\, model architectures\, or training procedures. More strikingly\, our further investigation implies that diffusion models are learning distinct distributions affected by the training data size and model capacity\, so that the model reproducibility manifests in two distinct training regimes with phase transition: (i) “memorization regime”\, where the diffusion model overfits to the training data distribution\, and (ii) “generalization regime”\, where the model learns the underlying data distribution and generate new samples with finite training data. Finally\, our results have strong practical implications regarding training efficiency\, model privacy\, and controllable generation of diffusion models\, and our work raises numerous intriguing theoretical questions for future investigation. \nBio: “Qing Qu is an assistant professor in EECS department at the University of Michigan. Prior to that\, he was a Moore-Sloan data science fellow at Center for Data Science\, New York University\, from 2018 to 2020. He received his Ph.D from Columbia University in Electrical Engineering in Oct. 2018. He received his B.Eng. from Tsinghua University in Jul. 2011\, and a M.Sc.from the Johns Hopkins University in Dec. 2012\, both in Electrical and Computer Engineering. His research interest lies at the intersection of foundation of data science\, machine learning\, numerical optimization\, and signal/image processing\, with focus on developing efficient nonconvex methods and global optimality guarantees for solving representation learning and nonlinear inverse problems in engineering and imaging sciences.\nHe is the recipient of Best Student Paper Award at SPARS’15\, and the recipient of Microsoft PhD Fellowship in machine learning in 2016\, and best paper awards in NeurIPS Diffusion Model Workshop in 2023. He received the NSF Career Award in 2022\, and Amazon Research Award (AWS AI) in 2023. He is the program chair of the new Conference on Parsimony & Learning.” URL:https://datascience.ucsd.edu/event/special-seminar-qing-qu/ LOCATION:Halıcıoğlu Data Science Institute (HDSI)\, Room 123\, 3234 Matthews Ln\, La Jolla\, CA\, 92093\, United States CATEGORIES:Seminar ATTACH;FMTTYPE=image/png:https://datascience.ucsd.edu/wp-content/uploads/2024/01/HDSI-UCSD-Image_Dark-blue-e1710178042629.png END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240327T140000 DTEND;TZID=America/Los_Angeles:20240327T153000 DTSTAMP:20260531T085249 CREATED:20240313T191359Z LAST-MODIFIED:20240323T081955Z UID:10000458-1711548000-1711553400@datascience.ucsd.edu SUMMARY:Towards a Machine Capable of Learning Everything | Hao Liu DESCRIPTION:Abstract: Large generative models such as ChatGPT have led to amazing results and revolutionized artificial intelligence. In this talk\, I will discuss my research on advancing the foundation of these models\, centered around addressing the architectural bottlenecks of learning from everything. First\, I will describe our efforts to remove context size limitations of the transformer architecture. Our new model architecture and training method allow for nearly infinitely large context sizes without approximations. Our proposed technique has been used for building state-of-the-art open-source and proprietary models. I will then discuss the applications of large context in world model learning and in reinforcement learning\, including Large World Model\, the world’s first multimodal model of million-length scale\, and the required training methodologies. Next\, I will introduce my research on unsupervised exploration that pioneered learning beyond existing knowledge\, allowing unsupervised pretrained models to outperform human experts in gameplay and paving the road for learning beyond imitating existing knowledge. Finally\, I will envision the modeling and training paradigms for the next generation of large generative models we should build\, focusing on advances in neural net architecture\, efficient scaling\, large context reasoning\, and discovery.” \nBio: Hao Liu is a final-year Ph.D. candidate in the Department of Electrical Engineering and Computer Sciences at UC Berkeley\, where he is advised by Pieter Abbeel. During his PhD\, he has also spent two years part-time at Google Brain and DeepMind. His research interests focus on the foundations of generative models\, including machine learning and neural networks\, with the goal of developing computationally scalable solutions for generalization. He recently developed Large World Model (LWM) and architectural advances (BlockwiseTransformers\, and RingAttention) for scaling transformers. Earlier\, he pioneered general and scalable unsupervised exploration (APT and APS). His work on million-length contexts has been influential at Google\, Meta\, and the broader industry. Several of his papers have been presented as spotlight and oral presentations at top-tier machine learning conferences\, and have also been featured in popular media\, including MarkTechPost\, Business Insider\, and ZDNet. URL:https://datascience.ucsd.edu/event/special-seminar-hao-liu/ LOCATION:Halıcıoğlu Data Science Institute (HDSI)\, Room 123\, 3234 Matthews Ln\, La Jolla\, CA\, 92093\, United States CATEGORIES:Seminar END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240326T140000 DTEND;TZID=America/Los_Angeles:20240326T153000 DTSTAMP:20260531T085249 CREATED:20240323T081729Z LAST-MODIFIED:20240323T081729Z UID:10000463-1711461600-1711467000@datascience.ucsd.edu SUMMARY:Making machine learning predictably reliable | Andrew Ilyas DESCRIPTION:Abstract: “Despite ML models’ impressive performance\, training and deploying them is currently a somewhat messy endeavor. But does it have to be? In this talk\, I overview my work on making ML “predictably reliable”—enabling developers to know when their models will work\, when they will fail\, and why. \nTo begin\, we use a case study of adversarial inputs to show that human intuition can be a poor predictor of how ML models operate. Motivated by this\, we present a line of work that aims to develop a precise understanding of the ML pipeline\, combining statistical tools with large-scale experiments to characterize the role of each individual design choice: from how to collect data\, to what dataset to train on\, to what learning algorithm to use.” \nBio: “Andrew Ilyas is a PhD student in Computer Science at MIT\, where he is advised by Aleksander Madry and Constantinos Daskalakis. His research aims to improve the reliability and predictability of machine learning systems. He was previously supported by an Open Philanthropy AI Fellowship.” URL:https://datascience.ucsd.edu/event/making-machine-learning-predictably-reliable-andrew-ilyas/ LOCATION:Halıcıoğlu Data Science Institute (HDSI)\, Room 123\, 3234 Matthews Ln\, La Jolla\, CA\, 92093\, United States END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240326T140000 DTEND;TZID=America/Los_Angeles:20240326T153000 DTSTAMP:20260531T085249 CREATED:20240304T171618Z LAST-MODIFIED:20240326T030740Z UID:10000451-1711461600-1711467000@datascience.ucsd.edu SUMMARY:Making machine learning predictably reliable | Andrew Ilyas DESCRIPTION:Abstract: “Despite ML models’ impressive performance\, training and deploying them is currently a somewhat messy endeavor. But does it have to be? In this talk\, I overview my work on making ML “predictably reliable”—enabling developers to know when their models will work\, when they will fail\, and why. \nTo begin\, we use a case study of adversarial inputs to show that human intuition can be a poor predictor of how ML models operate. Motivated by this\, we present a line of work that aims to develop a precise understanding of the ML pipeline\, combining statistical tools with large-scale experiments to characterize the role of each individual design choice: from how to collect data\, to what dataset to train on\, to what learning algorithm to use.” \n\nBio “Andrew Ilyas is a PhD student in Computer Science at MIT\, where he is advised by Aleksander Madry and Constantinos Daskalakis. His research aims to improve the reliability and predictability of machine learning systems. He was previously supported by an Open Philanthropy AI Fellowship.” URL:https://datascience.ucsd.edu/event/special-seminar-andrew-ilyas/ LOCATION:Halıcıoğlu Data Science Institute (HDSI)\, Room 123\, 3234 Matthews Ln\, La Jolla\, CA\, 92093\, United States CATEGORIES:Seminar ATTACH;FMTTYPE=image/png:https://datascience.ucsd.edu/wp-content/uploads/2024/01/HDSI-UCSD-Image_Dark-blue-e1710178042629.png END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240322T140000 DTEND;TZID=America/Los_Angeles:20240322T153000 DTSTAMP:20260531T085249 CREATED:20240313T190501Z LAST-MODIFIED:20240321T174552Z UID:10000457-1711116000-1711121400@datascience.ucsd.edu SUMMARY:Efficient Deep Learning with Sparsity: Algorithms\, Systems\, and Applications | Zhijian Liu DESCRIPTION:Abstract: Deep learning is used across a broad spectrum of applications. However\, behind its remarkable performance lies an increasing gap between the demand for and supply of computation. On the demand side\, the computational costs of deep learning models have surged dramatically\, driven by ever-larger input and model sizes. On the supply side\, as Moore’s Law slows down\, hardware no longer delivers increasing performance within the same power budget. \nIn this talk\, I will discuss my research efforts to bridge this demand-supply gap through the lens of sparsity. I will begin with my research on input sparsity. First\, I will introduce algorithms that systematically eliminate the least important patches/tokens from dense input data\, such as images\, enabling up to 60% sparsity without any loss in accuracy. Then\, I will present the system library that we have developed to effectively translate the theoretical savings from sparsity to practical speedups on hardware. Our system is up to 3 times faster than the leading industry solution from NVIDIA. Following this\, I will touch on my research on model sparsity\, highlighting a family of automated\, hardware-aware model compression frameworks that surpass manual solutions in accuracy and reduce the design cycle from weeks of human efforts to mere hours of GPU computation. Finally\, I will demonstrate the use of sparsity to accelerate a wide range of computation-intensive AI applications\, such as autonomous driving\, language modeling\, and high-energy physics. I will conclude this talk with my vision towards building more efficient and accessible AI. \nBio: Zhijian Liu is a Ph.D. candidate at MIT\, advised by Song Han. His research focuses on efficient machine learning and systems. He has developed efficient ML algorithms and provided them with effective system support. He has also contributed to accelerating computation-intensive AI applications in computer vision\, natural language processing\, and scientific discovery. His work has been featured as oral and spotlight presentations at conferences such as NeurIPS\, ICLR\, and CVPR. He was selected as the recipient of the Qualcomm Innovation Fellowship and the NVIDIA Graduate Fellowship. He was also recognized as a Rising Star in ML and Systems by MLCommons and a Rising Star in Data Science by UChicago and UCSD. Previously\, he was the founding research scientist at OmniML\, which was acquired by NVIDIA. URL:https://datascience.ucsd.edu/event/special-seminar-zhijian-liu/ LOCATION:Halıcıoğlu Data Science Institute (HDSI)\, Room 123\, 3234 Matthews Ln\, La Jolla\, CA\, 92093\, United States CATEGORIES:Seminar ATTACH;FMTTYPE=image/png:https://datascience.ucsd.edu/wp-content/uploads/2024/01/HDSI-UCSD-Image_Dark-blue-e1710178042629.png END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240320T140000 DTEND;TZID=America/Los_Angeles:20240320T153000 DTSTAMP:20260531T085249 CREATED:20240313T184800Z LAST-MODIFIED:20240318T230405Z UID:10000456-1710943200-1710948600@datascience.ucsd.edu SUMMARY:Understanding Deep Learning through Optimization Geometry| Nati (Nathan) Srebro DESCRIPTION:Abstract: How can models with more parameters than training examples generalize well\, and generalize even better when we add even more parameters\, even without explicit complexity control? In recent years\, it is becoming increasingly clear that much\, or perhaps all\, of the complexity control and generalization ability of deep learning comes from the optimization bias\, or implicit bias\, of the training procedures. In this talk\, I will survey our work from the past several years on highlighting the role of optimization geometry in determining such implicit bias\, and understanding deep learning through it\, and how this view influences the study of further deep learning phenomena. \nBio: Nati (Nathan) Srebro is a professor at the Toyota Technological Institute at Chicago\, with cross-appointments at the University of Chicago’s Department of Computer Science\, and Committee on Computational and Applied Mathematics. He obtained his PhD from the Massachusetts Institute of Technology in 2004\, and previously was a postdoctoral fellow at the University of Toronto\, a visiting scientist at IBM\, and an associate professor at the Technion\, and held visiting position at the Weizmann Institute and at École Polytechnique Fédérale de Lausanne. \nDr. Srebro’s research encompasses methodological\, statistical and computational aspects of machine learning\, as well as related problems in optimization. Some of Srebro’s significant contributions include work on learning “wider” Markov networks\, introducing the use of the nuclear norm for machine learning\, introducing the “equalized odds” fairness notion for non-discrimination\, work on fast optimization techniques for machine learning\, and on the relationship between learning and optimization. \nWebsite: https://nati.ttic.edu/ URL:https://datascience.ucsd.edu/event/special-seminar-nathan-srebro/ LOCATION:Halıcıoğlu Data Science Institute (HDSI)\, Room 123\, 3234 Matthews Ln\, La Jolla\, CA\, 92093\, United States CATEGORIES:Seminar ATTACH;FMTTYPE=image/png:https://datascience.ucsd.edu/wp-content/uploads/2024/01/HDSI-UCSD-Image_Dark-blue-e1710178042629.png END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240320T100000 DTEND;TZID=America/Los_Angeles:20240320T110000 DTSTAMP:20260531T085249 CREATED:20240318T224758Z LAST-MODIFIED:20240318T224925Z UID:10000461-1710928800-1710932400@datascience.ucsd.edu SUMMARY:TILOS Seminar: How Large Models of Language and Vision Help Agents to Learn to Behave DESCRIPTION:Roy Fox\, Assistant Professor and Director of the Intelligent Dynamics Lab at UC Irvine\nHDSI 123 and Zoom (Link below) \nAbstract: If learning from data is valuable\, can learning from big data be very valuable? So far\, it has been so in vision and language\, for which foundation models can be trained on web-scale data to support a plethora of downstream tasks; not so much in control\, for which scalable learning remains elusive. Can information encoded in vision and language models guide reinforcement learning of control policies? In this talk\, I will discuss several ways for foundation models to help agents to learn to behave. Language models can provide better context for decision-making: we will see how they can succinctly describe the world state to focus the agent on relevant features; and how they can form generalizable skills that identify key subgoals. Vision and vision–language models can help the agent to model the world: we will see how they can block visual distractions to keep state representations task-relevant; and how they can hypothesize about abstract world models that guide exploration and planning. \nBio: Roy Fox is an Assistant Professor of Computer Science at the University of California\, Irvine. His research interests include theory and applications of control learning: reinforcement learning (RL)\, control theory\, information theory\, and robotics. His current research focuses on structured and model-based RL\, language for RL and RL for language\, and optimization in deep control learning of virtual and physical agents. URL:https://datascience.ucsd.edu/event/tilos-seminar-how-large-models-of-language-and-vision-help-agents-to-learn-to-behave/ LOCATION:Halıcıoğlu Data Science Institute (HDSI)\, Room 123\, 3234 Matthews Ln\, La Jolla\, CA\, 92093\, United States CATEGORIES:Seminar ATTACH;FMTTYPE=image/png:https://datascience.ucsd.edu/wp-content/uploads/2023/10/TILOS-Square_HDSI-Website-e1712854679822.png END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240315T110000 DTEND;TZID=America/Los_Angeles:20240315T143000 DTSTAMP:20260531T085249 CREATED:20240122T000022Z LAST-MODIFIED:20240315T011310Z UID:10000428-1710500400-1710513000@datascience.ucsd.edu SUMMARY:HDSI Capstone Showcase DESCRIPTION: URL:https://dsc-capstone.org/showcase-24/ LOCATION:Price Center East Ballroom\, 9500 Gilman Drive\, La Jolla\, CA\, 92093\, United States CATEGORIES:Student Event ATTACH;FMTTYPE=image/png:https://datascience.ucsd.edu/wp-content/uploads/2024/01/HDSI-UCSD-Image-1.png END:VEVENT BEGIN:VEVENT DTSTART;VALUE=DATE:20240315 DTEND;VALUE=DATE:20240317 DTSTAMP:20260531T085249 CREATED:20240307T172716Z LAST-MODIFIED:20240307T172716Z UID:10000454-1710460800-1710633599@datascience.ucsd.edu SUMMARY:HDSI LLM Workshop DESCRIPTION: URL:https://ucsd-hdsi-llm-workshop.github.io/2024/index.html LOCATION:Halıcıoğlu Data Science Institute (HDSI)\, Room 123\, 3234 Matthews Ln\, La Jolla\, CA\, 92093\, United States CATEGORIES:Workshops ATTACH;FMTTYPE=image/png:https://datascience.ucsd.edu/wp-content/uploads/2024/01/HDSI-UCSD-Image-e1712856546428.png END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240313T130000 DTEND;TZID=America/Los_Angeles:20240313T140000 DTSTAMP:20260531T085249 CREATED:20240313T195909Z LAST-MODIFIED:20240313T195909Z UID:10000460-1710334800-1710338400@datascience.ucsd.edu SUMMARY:Domain Counterfactuals for Trustworthy ML via Sparse Interventions | David I. Inouye DESCRIPTION:Talk Abstract: \nAlthough incorporating causal concepts into deep learning shows promise for increasing explainability\, fairness\, and robustness\, existing methods require unrealistic assumptions and aim to recover the full latent causal model. This talk proposes an alternative: domain counterfactuals. Domain counterfactuals ask a more concrete question: “What would a sample look like if it had been generated in a different domain (or environment)?”   This avoids the challenges of full causal recovery while answering an important causal query. I will theoretically analyze the domain counterfactual problem for invertible causal models and prove an estimation bound that depends on the sparsity of intervention\, i.e.\, the number of intervened causal variables.  Leveraging this theory\, I will introduce a practical counterfactual estimation algorithm that outperforms baselines. Additionally\, I will showcase the potential of domain counterfactuals for counterfactual fairness and domain generalization through preliminary results. Finally\, I will connect this work to my broader research focus on distribution matching\, highlighting its potential as a foundational tool for building trustworthy machine learning systems. \nBio: \nProf. David I. Inouye is an assistant professor in the Elmore Family School of Electrical and Computer Engineering at Purdue University. His lab focuses on trustworthy machine learning (ML)\, which aims to make ML systems more robust\, causal and explainable. Currently\, he is interested in advancing distribution matching algorithms and applications such as causality\, domain generalization\, and distribution shift explanations. He is also interested in highly robust distributed learning algorithms on a network of devices\, called Internet Learning. His research is funded by ARL\, ONR\, and NSF. Previously\, he was a postdoc at Carnegie Mellon University working with Prof. Pradeep Ravikumar. He completed his Computer Science PhD at The University of Texas at Austin in 2017 advised by Prof. Inderjit Dhillon and Prof. Pradeep Ravikumar. He was awarded the NSF Graduate Research Fellowship (NSF GRFP). URL:https://datascience.ucsd.edu/event/domain-counterfactuals-for-trustworthy-ml-via-sparse-interventions-david-i-inouye/ LOCATION:Halıcıoğlu Data Science Institute (HDSI)\, Room 404\, 3234 Matthews Ln\, La Jolla\, CA\, 92093\, United States CATEGORIES:Seminar END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240312T140000 DTEND;TZID=America/Los_Angeles:20240312T153000 DTSTAMP:20260531T085249 CREATED:20240304T171426Z LAST-MODIFIED:20240311T181154Z UID:10000450-1710252000-1710257400@datascience.ucsd.edu SUMMARY:From Pixels to Measurements: Understanding the Dynamic World ~ Adam Harley DESCRIPTION:In computer vision\, “video understanding” typically concerns summarization: tracking the main objects\, or describing the main actions. While progress here has been impressive\, many practical applications require extracting information which is much more fine-grained. For example\, biologists are highly interested in tracking specific key points of organisms in long video recordings. Algorithms for such tasks require the generality and precision of low-level vision methods (e.g.\, optical flow)\, but benefit from knowledge about the physical world (e.g.\, things continue to exist while they are occluded). In this talk\, I will present our progress on this crucial space of problems. Our central contribution is to widen the window of “temporal context” used for inference: instead of tracking entities from one frame to the next\, we inspect dozens of frames simultaneously\, and return an answer that makes sense for the full clip. I will discuss the methods and datasets that we have created to drive progress along these lines\, and highlight natural science applications of the work. Finally\, I will introduce our ongoing effort to produce a “foundation model” of motion\, aiming to deliver arbitrary-granularity tracking for a huge variety of real-world situations. \nAdam is a postdoctoral scholar at Stanford University\, working with Leonidas Guibas. He received a Ph.D. in robotics from Carnegie Mellon University\, where he worked with Katerina Fragkiadaki. He received his M.S. in Computer Science at Toronto Metropolitan University\, working with Kosta Derpanis. Adam is a recipient of the NSERC PGS-D scholarship\, and the Toronto Metropolitan University Gold Medal. His research interests lie in Computer Vision and Machine Learning\, particularly for 3D understanding and fine-grained tracking. URL:https://datascience.ucsd.edu/event/special-seminar-adam-harley/ LOCATION:Halıcıoğlu Data Science Institute (HDSI)\, Room 123\, 3234 Matthews Ln\, La Jolla\, CA\, 92093\, United States CATEGORIES:Seminar ATTACH;FMTTYPE=image/png:https://datascience.ucsd.edu/wp-content/uploads/2024/01/HDSI-UCSD-Image_Dark-blue-e1710178042629.png END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240311T140000 DTEND;TZID=America/Los_Angeles:20240311T153000 DTSTAMP:20260531T085249 CREATED:20240304T171239Z LAST-MODIFIED:20240304T171239Z UID:10000449-1710165600-1710171000@datascience.ucsd.edu SUMMARY:Special Seminar | Zhuang Liu DESCRIPTION:Talk info: to be provided URL:https://datascience.ucsd.edu/event/special-seminar-zhuang-liu/ LOCATION:Halıcıoğlu Data Science Institute (HDSI)\, Room 123\, 3234 Matthews Ln\, La Jolla\, CA\, 92093\, United States CATEGORIES:Seminar ATTACH;FMTTYPE=image/png:https://datascience.ucsd.edu/wp-content/uploads/2024/01/HDSI-UCSD-Image_Dark-blue-e1710178042629.png END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240308T120000 DTEND;TZID=America/Los_Angeles:20240308T130000 DTSTAMP:20260531T085249 CREATED:20240226T212933Z LAST-MODIFIED:20240318T225308Z UID:10000447-1709899200-1709902800@datascience.ucsd.edu SUMMARY:TILOS Webinar: AI Ethics in Research DESCRIPTION:The Ethics and Early Career Committee would like to invite you to our upcoming webinar on AI Ethics in Research. This will take place virtually through Zoom on Friday\, March 8th at noon Pacific\, 2pm Central\, 3pm Eastern (https://nu.zoom.us/j/2183621123). \nPlease join Dr. Nisheeth Vishnoi from Yale and Dr. David Danks from UC San Diego who will discuss their Research in AI Ethics. Professor Danks develops practical frameworks and methods to incorporate ethical and policy considerations throughout the AI lifecycle\, including different ways to include them in optimization steps. Bias and fairness have been a particular focus given the multiple ways in which they can be measured\, represented\, and used. Professor Vishnoi uses optimization as a lens to study how subjective human and societal biases emerge in the objective world of artificial algorithms\, as well as how to design strategies to mitigate these biases. \nThis event is a great opportunity to learn about the constantly evolving issues of AI Ethics in research and the societal impact of AI. It will also provide a platform for students to gain insights and valuable advice that can help them in their future career pursuits. \nNisheeth Vishnoi is the A. Bartlett Giamatti Professor of Computer Science and a co-founder of the Computation and Society Initiative at Yale University. He studies the foundations of computation\, and his research spans several areas of theoretical computer science\, optimization\, and machine learning.  He is also interested in understanding nature and society from a computational viewpoint. Here\, his current focus includes understanding the emergence of intelligence and developing methods to address ethical issues at the interface of artificial intelligence and humanity. \nDavid Danks is Professor of Data Science & Philosophy and affiliate faculty in Computer Science & Engineering at University of California\, San Diego. His research interests range widely across philosophy\, cognitive science\, and machine learning\, including their intersection. Danks has examined the ethical\, psychological\, and policy issues around AI and robotics across multiple sectors\, including transportation\, healthcare\, privacy\, and security. He has also done significant research in computational cognitive science and developed multiple novel causal discovery algorithms for complex types of observational and experimental data. Danks is the recipient of a James S. McDonnell Foundation Scholar Award\, as well as an Andrew Carnegie Fellowship. He currently serves on multiple advisory boards\, including the National AI Advisory Committee. URL:https://datascience.ucsd.edu/event/tilos-webinar-ai-ethics-in-research/ LOCATION:Virtual CATEGORIES:Webinar ATTACH;FMTTYPE=image/png:https://datascience.ucsd.edu/wp-content/uploads/2023/10/TILOS-Square_HDSI-Website-e1712854679822.png END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240306T170000 DTEND;TZID=America/Los_Angeles:20240306T183000 DTSTAMP:20260531T085249 CREATED:20240209T162735Z LAST-MODIFIED:20240209T163214Z UID:10000438-1709744400-1709749800@datascience.ucsd.edu SUMMARY:The Ethical and Policy Implications of Artificial Intelligence DESCRIPTION:The Institute for Practical Ethics welcomes David Danks as the 2024 keynote speaker. \nDanks\, a UC San Diego professor in the Department of Philosophy and Halıcıoğlu Data Science Institute\, is an expert researcher at the intersection of philosophy\, cognitive science and machine learning. He serves on multiple boards\, including the United States National AI Advisory Committee. \nArtificial intelligence is seemingly everywhere today\, both in public perception and in our everyday lives. This growth has led to many stories about the widespread harms that can result from AI done poorly. As a result\, there are now numerous demands for ‘ethical AI\,’ but relatively little understanding of what that might involve. \nIn this keynote\, David Danks will explore the nature of responsible AI\, arguing that it involves much more than code or data. He will critically assess current approaches to producing more responsible AI\, then suggest key policy and practical approaches that would likely be more effective. It is critical we create more responsible AI\, but that will require rethinking many of our current practices in academia\, government and industry. URL:https://www.eventbrite.com/e/the-ethical-and-policy-implications-of-artificial-intelligence-tickets-817599541237?aff=ipewebsite LOCATION:Sanford Consortium CATEGORIES:Seminar ATTACH;FMTTYPE=image/jpeg:https://datascience.ucsd.edu/wp-content/uploads/2024/02/IPE_David-Danks.jpg END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240304T140000 DTEND;TZID=America/Los_Angeles:20240304T150000 DTSTAMP:20260531T085249 CREATED:20240304T170615Z LAST-MODIFIED:20240304T170615Z UID:10000448-1709560800-1709564400@datascience.ucsd.edu SUMMARY:Evaluating and Designing Computing Systems for the Future of Work | Hancheng Cao DESCRIPTION:Abstract: From collaborative software to generative AI\, computing technologies are redefining the way we work\, communicate and collaborate. Yet with the growing complexities of computing platforms\, it becomes increasingly challenging to foresee their impacts on human behavior\, leading to not only poor user experience but also problematic applications that mirror and amplify societal issues. How can we better understand machine behavior and machine-mediated user behavior over computing platforms? How can we build applications that align with our needs and values with emerging computing technologies? My research aims to answer these questions through novel measurements and computational methods inspired by social science insights\, such as mining increasingly available large-scale data on how people build\, adopt\, and interact with computing systems. In this talk\, I will present my work demonstrating this approach in the future of work context\, where I develop data-driven\, AI-powered and human-centered methods to understand\, evaluate and design sociotechnical systems at the workplace. I will present an analysis of remote meeting experience through mining millions of meetings\, a study on how an AI algorithm can be built to predict team fracture\, and a development and evaluation study on a generative AI-based scientific feedback system for researchers. These projects exemplify the opportunities to leverage computation and data to better understand\, support and augment work practices.         \nBio: Hancheng Cao is a final year PhD candidate in computer science (with a PhD minor in management science and engineering) at Stanford University working with Prof. Daniel McFarland and Prof. Michael Bernstein. He works in the field of computational social science and human computer interaction\, where he mines large-scale data\, develops algorithms and builds systems to study human behavior. Recognized as a Stanford Interdisciplinary Graduate Fellow\, he has published 30 academic papers across fields\, with three works he led recognized as Best Paper (CHI 2023) or Honorable Mention (CSCW 2020\, CHI 2021) awards. His research has also appeared in leading social science journals (e.g. American Sociological Review). His research has been widely covered in the media\, including Wired\, Forbes\, New Scientist\, TED among others. URL:https://datascience.ucsd.edu/event/evaluating-and-designing-computing-systems-for-the-future-of-work-hancheng-cao/ LOCATION:Halıcıoğlu Data Science Institute (HDSI)\, Room 123\, 3234 Matthews Ln\, La Jolla\, CA\, 92093\, United States ATTACH;FMTTYPE=image/png:https://datascience.ucsd.edu/wp-content/uploads/2024/01/HDSI-UCSD-Image_Dark-blue-e1710178042629.png END:VEVENT BEGIN:VEVENT DTSTART;VALUE=DATE:20240304 DTEND;VALUE=DATE:20240307 DTSTAMP:20260531T085249 CREATED:20240304T170710Z LAST-MODIFIED:20240304T171028Z UID:10000445-1709510400-1709769599@datascience.ucsd.edu SUMMARY:EnCORE Workshop | Old Questions and New Directions in Theory of Clustering DESCRIPTION:We are hosting an EnCORE workshop on Old Questions and New Directions in Theory of Clustering at UCSD from March 4th to 6th\, 2024. While in person registration is closed due to limited seats availability\, you can register to attend the workshop virtually here: https://sites.google.com/view/clusteringinsandiego \nWe have a stellar lineup of speakers and we hope the workshop will generate many new questions to work on. \n  URL:https://sites.google.com/view/clusteringinsandiego LOCATION:Virtual CATEGORIES:Workshops ATTACH;FMTTYPE=image/png:https://datascience.ucsd.edu/wp-content/uploads/2023/10/Encore-logo_HDSI-Website.png END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240226T123000 DTEND;TZID=America/Los_Angeles:20240226T140000 DTSTAMP:20260531T085249 CREATED:20240213T221414Z LAST-MODIFIED:20240220T235645Z UID:10000443-1708950600-1708956000@datascience.ucsd.edu SUMMARY:Building Human-AI Alignment: Specifying\, Inspecting\, and Modeling AI Behaviors | Serena Booth DESCRIPTION:Abstract: The learned behaviors of AI and robot agents should align with the intentions of their human designers. Alignment is necessary for AI systems to be used in many sectors of the economy\, and so the process of aligning AI systems becomes critical to study for defining effective AI policy. Toward this goal\, people must be able to easily specify\, inspect\, and model agent behaviors. For specifications\, we will consider expert-written reward functions for reinforcement learning (RL) and non-expert preferences for reinforcement learning from human feedback (RLHF). I will show evidence that experts are bad at writing reward functions: even in a trivial setting\, experts write specifications that are overfit to a particular RL algorithm\, and they often write erroneous specifications for agents that fail to encode their true intent. I will also show that the common approach to learning a reward function from non-experts in RLHF uses an inductive bias that fails to encode how humans express preferences\, and that our proposed bias better encodes human preferences both theoretically and empirically. I will discuss the policy implications: namely\, that engineers’ design processes and embedded assumptions in building AI must be considered. For inspection\, humans must be able to assess the behaviors an agent learns from a given specification. I will discuss a method to find settings that exhibit particular behaviors\, like out-of-distribution failures. I will discuss the policy implications for testing AI systems\, for example through red teaming. Lastly\, cognitive science theories attempt to show how people build conceptual models that explain agent behaviors. I will show evidence that some of these theories are used in research to support humans\, but that we can still build better curricula for modeling. I will discuss the policy need for careful onboarding to AI systems. I will end by discussing my current work in the U.S. Senate on responding to the proliferation of AI. Collectively\, my research provides evidence that—even with the best of intentions— current human-AI systems often fail to induce alignment\, and my research proposes promising directions for how to build better aligned human-AI systems. URL:https://datascience.ucsd.edu/event/special-seminar-serena-booth/ LOCATION:GPS\, Robinson Building Complex (RBC)\, 3106 CATEGORIES:Seminar END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240222T140000 DTEND;TZID=America/Los_Angeles:20240222T150000 DTSTAMP:20260531T085249 CREATED:20240126T183316Z LAST-MODIFIED:20240202T225717Z UID:10000431-1708610400-1708614000@datascience.ucsd.edu SUMMARY:The continuum of gene regulation at single cell resolution\, from Drosophila development to human complex traits | Diego Calderon DESCRIPTION:Single-cell technologies have emerged as powerful tools for studying development\, enabling comprehensive surveys of cellular diversity at profiled timepoints. They shed light on the dynamics of regulatory element activity and gene expression changes during the emergence of each cell type. Despite their potential\, nearly all atlases of embryogenesis are constrained by sampling density\, i.e.\, the number of discrete time points at which individual embryos are harvested. This limitation affects the resolution at which regulatory transitions can be characterized. In this talk\, I present a novel cell collection approach capable of constructing a continuous representation of dynamic regulatory processes. I applied this approach to generate a continuous\, single-cell atlas of chromatin accessibility and gene expression spanning Drosophila embryogenesis. Additionally\, I will discuss my past and future research\, applying new genomic technologies to characterize gene regulation important for human diseases. URL:https://datascience.ucsd.edu/event/special-seminar-diego-calderon/ LOCATION:Powell-Focht Bioengineering Hall (PFBH)\, FUNG Auditorium CATEGORIES:Seminar ATTACH;FMTTYPE=image/png:https://datascience.ucsd.edu/wp-content/uploads/2024/01/HDSI-UCSD-Image_Dark-blue-e1710178042629.png END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240221T140000 DTEND;TZID=America/Los_Angeles:20240221T153000 DTSTAMP:20260531T085249 CREATED:20240205T171619Z LAST-MODIFIED:20240220T172031Z UID:10000436-1708524000-1708529400@datascience.ucsd.edu SUMMARY:The Synergy between Machine Learning and the Natural Sciences | Max Welling DESCRIPTION:Abstract: Traditionally machine learning has been heavily influenced by neuroscience (hence the name artificial neural networks) and physics (e.g. MCMC\, Belief Propagation\, and Diffusion based Generative AI). We have recently witnessed that the flow of information has also reversed\, with new tools developed in the ML community impacting physics\, chemistry and biology. Examples include faster DFT\, Force-Field accelerated MD simulations\, PDE Neural Surrogate models\, generating druglike molecules\, and many more. In this talk I will review the exciting opportunities for further cross fertilization between these fields\, ranging from faster (classical) DFT calculations and enhanced transition path sampling to traveling waves in artificial neural networks. \nBio: Prof. Max Welling is a research chair in Machine Learning at the University of Amsterdam and a Distinguished Scientist at MSR. He is a fellow at the Canadian Institute for Advanced Research (CIFAR) and the European Lab for Learning and Intelligent Systems (ELLIS) where he also serves on the founding board. His previous appointments include VP at Qualcomm Technologies\, professor at UC Irvine\, postdoc at U. Toronto and UCL under supervision of prof. Geoffrey Hinton\, and postdoc at Caltech under supervision of prof. Pietro Perona. He finished his PhD in theoretical high energy physics under supervision of Nobel laureate prof. Gerard ‘t Hooft. \nMax Welling has served as associate editor in chief of IEEE TPAMI from 2011-2015\, he serves on the advisory board of the Neurips foundation since 2015 and has been program chair and general chair of Neurips in 2013 and 2014 respectively. He was also program chair of AISTATS in 2009 and ECCV in 2016 and general chair of MIDL 2018. Max Welling is recipient of the ECCV Koenderink Prize in 2010 and the ICML Test of Time award in 2021. He directs the Amsterdam Machine Learning Lab (AMLAB) and co-directs the Qualcomm-UvA deep learning lab (QUVA) and the Bosch-UvA Deep Learning lab (DELTA). URL:https://datascience.ucsd.edu/event/distinguished-colloquium-max-welling/ LOCATION:Halıcıoğlu Data Science Institute (HDSI)\, Room 123\, 3234 Matthews Ln\, La Jolla\, CA\, 92093\, United States CATEGORIES:Guest Lecture ATTACH;FMTTYPE=image/png:https://datascience.ucsd.edu/wp-content/uploads/2024/02/Max_Welling_DLS_1240x650.png END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240221T123000 DTEND;TZID=America/Los_Angeles:20240221T140000 DTSTAMP:20260531T085249 CREATED:20240220T163858Z LAST-MODIFIED:20240220T163858Z UID:10000444-1708518600-1708524000@datascience.ucsd.edu SUMMARY:Computational approaches for uncovering implicit strategies in political discourse | Julia Mendelsohn DESCRIPTION:When discussing politics\, people often use subtle linguistic strategies to influence how their audience thinks about issues\, which can then impact public opinion and policy. For example\, anti-immigration activists may frame immigration as a threat to native born citizens’ jobs\, describe immigrants with dehumanizing vermin-related metaphors\, or even use coded expressions to covertly connect immigration with antisemitic conspiracy theories. This talk will focus on the development of computational approaches to analyze three strategies: framing\, dehumanization\, and dogwhistle communication. I will discuss how I draw from multiple social science disciplines to develop typologies and curate data resources\, as well as how I build and evaluate natural language processing models for detecting these strategies. I further analyze the use of these strategies in political discourse across several domains\, and assess the implications of such nuanced rhetoric for both society and technology. URL:https://datascience.ucsd.edu/event/computational-approaches-for-uncovering-implicit-strategies-in-political-discourse-julia-mendelsohn-2/ LOCATION:GPS\, Robinson Building Complex (RBC)\, 3106 CATEGORIES:Seminar ATTACH;FMTTYPE=image/png:https://datascience.ucsd.edu/wp-content/uploads/2024/01/HDSI-UCSD-Image_Dark-blue-e1710178042629.png END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240220T130000 DTEND;TZID=America/Los_Angeles:20240220T143000 DTSTAMP:20260531T085249 CREATED:20240220T182547Z LAST-MODIFIED:20240220T182547Z UID:10000446-1708434000-1708439400@datascience.ucsd.edu SUMMARY:Learning Inductive Representations for Reasoning over Knowledge Graphs | Zhaocheng Zhu DESCRIPTION:Abstract: Reasoning\, the ability to logically draw conclusions from existing knowledge\, has been long pursued as a goal of artificial intelligence. Although numerous learning algorithms have been developed for reasoning\, most of them are limited to the domain they are trained on. By contrast\, humans often derive high-level rules or principles from experience and apply them to new domains — an ability referred as inductive generalization. In this talk\, we present a series of works that learn inductive representations for reasoning over knowledge graphs. First\, we introduce Neural Bellman-Ford Networks (NBFNet) that captures paths between entities and can generalize to graphs of new entities. Then we discuss Graph Neural Network Query Executor (GNN-QE)\, an extension of NBNet that answers multi-hop logical queries and generalizes well on our inductive benchmark. Finally\, by learning inductive representations for both entities and relations\, we demonstrate that a model can generalize to any graph with arbitrary entity and relation vocabularies\, paving the way for foundation models for knowledge graph reasoning. \n \nBio: Zhaocheng Zhu is a final-year Ph.D. candidate advised by Prof. Jian Tang at Mila – Quebec AI Institute\, University of Montreal. His research interests include reasoning\, knowledge graphs and large language models. His works\, among the first to study inductive generalization across structures\, have led to a paradigm shift away from traditional knowledge graph embedding methods that have been used for years. He gave a tutorial on knowledge graph reasoning at AAAI 2022. He is also an active developer of machine learning systems\, and led the development of two open-source libraries\, GraphVite for large-scale embedding training and TorchDrug for drug discovery research. URL:https://datascience.ucsd.edu/event/learning-inductive-representations-for-reasoning-over-knowledge-graphs-zhaocheng-zhu/ LOCATION:Computer Science & Engineering Building (CSE)\, Room 1242\, 3234 Matthews Ln\, La Jolla\, CA\, 92093\, United States CATEGORIES:Seminar END:VEVENT BEGIN:VEVENT DTSTART;TZID=America/Los_Angeles:20240215T123000 DTEND;TZID=America/Los_Angeles:20240215T140000 DTSTAMP:20260531T085249 CREATED:20240213T220919Z LAST-MODIFIED:20240213T221018Z UID:10000441-1708000200-1708005600@datascience.ucsd.edu SUMMARY:Targeting humanitarian aid with machine learning and digital data | Emily Aiken DESCRIPTION:Abstract: The majority of humanitarian aid and social protection programs globally are targeted\, providing assistance to individuals or communities identified to be poorest or most in need. In low- and middle-income countries\, the targeting of aid programs is often limited by low-quality\, out-of-date\, or missing data on poverty and vulnerability. Novel “big” digital data sources\, such as those captured by satellites\, mobile phones\, and financial services providers — when combined with advances in machine learning — can improve the accuracy of aid program targeting. In this talk\, I will cover empirical results on the accuracy of these new data-driven and algorithmic approaches to aid allocation\, and will discuss emergent implications for fairness\, privacy\, transparency\, and community dynamics. URL:https://datascience.ucsd.edu/event/targeting-humanitarian-aid-with-machine-learning-and-digital-data-emily-aiken/ LOCATION:GPS\, Robinson Building Complex (RBC)\, 3203 CATEGORIES:Seminar END:VEVENT END:VCALENDAR