BEGIN:VCALENDAR VERSION:2.0 PRODID:-//CalDiscovery//caldiscovery//EN CALSCALE:GREGORIAN METHOD:PUBLISH DESCRIPTION:Discovered and maintained by CalDiscovery (https://caldiscover y.com). Event details belong to their publishers. Contains information fro m CalDiscovery\, which is made available under the ODbL 1.0. SOURCE:https://caldiscovery.com/anaheim-ca/src-10359.ics X-PUBLISHED-TTL:PT1H X-WR-CALDESC:Discovered and maintained by CalDiscovery (https://caldiscove ry.com). Event details belong to their publishers. Contains information fr om CalDiscovery\, which is made available under the ODbL 1.0. X-WR-CALNAME:CalDiscovery - anaheim-ca BEGIN:VEVENT SUMMARY:San Diego (Leidos) ICS Alumni Reception 2026 DTSTART:20260918T003000Z DTEND:20260918T023000Z DTSTAMP:20260903T180718Z UID:10001091-1789666200-1789673400@ics.uci.edu CATEGORIES:community DESCRIPTION:We’re excited to return to San Diego for our annual ICS Alum ni Reception! \nYou and a guest are cordially invited to join Dean Marios Papaefthymiou and fellow ICS alumni at this special alumni reception in Sa n Diego\, CA. \nWhen: Thursday\, September 17\, 2026\nTime: 5:30PM – 7:3 0PM \nLocation: Leidos\n4155 Campus Point Ct\, San Diego\, CA 92121 \nHost ed on the beautiful rooftop at Leidos\, the evening brings together fellow Anteaters\, ICS Dean Marios Papaefthymiou\, and members of the ICS commun ity for savory fare\, cocktails and meaningful connections. \nPlease RSVP by Monday\, September 14. LAST-MODIFIED:20260903T180718Z URL:https://ics.uci.edu/event/san-diego-leidos-ics-alumni-reception-2026/ X-SOURCE-NAME:UC Irvine Donald Bren School of Information & Computer Scien ces X-SOURCE-URL:https://ics.uci.edu/event/seattle-area-ics-alumni-reception-2 026 END:VEVENT BEGIN:VEVENT SUMMARY:Fixed Point Computation Problems and Facets of Complexity DTSTART:20260925T180000Z DTEND:20260925T190000Z DTSTAMP:20260729T211044Z UID:10001067-1790334000-1790337600@ics.uci.edu CATEGORIES:education DESCRIPTION:Abstract: Many problems from a wide variety of areas can be fo rmulated mathematically as the problem of computing a fixed point of a sui table given multivariate function. Examples include a variety of problems from game theory\, economics\, optimization\, stochastic analysis\, verifi cation\, and others. In some problems there is a unique fixed point (for e xample if the function is a contraction)\; in others there may be multiple fixed points and any one of them is an acceptable solution\; while in oth er cases the desired object is a specific fixed point (for example the lea st fixed point or greatest fixed point of a monotone function). In this ta lk we will discuss several types of fixed point computation problems\, the ir complexity\, and some of the common themes that have emerged: classes o f problems for which there are efficient algorithms\, and other classes fo r which there seem to be serious obstacles. \nBio: Mihalis Yannakakis is t he Percy K. and Vida L. W. Hudson Professor of Computer Science at Columbi a University. Prior to joining Columbia\, he was Head of the Computing Pri nciples Research Department at Bell Labs\, and Professor of Computer Scien ce at Stanford University. Dr. Yannakakis received his PhD from Princeton University. He has served on the editorial boards of several journals\, in cluding as the editor-in-chief of the SIAM Journal on Computing\, and has chaired various conferences\, including STOC\, FOCS and PODS. Dr. Yannakak is is a recipient of the Knuth Prize\, the INFORMS John von Neumann Theory Prize\, he is a member of the National Academy of Sciences\, the National Academy of Engineering\, the American Academy of Arts and Sciences\, and of Academia Europaea. LAST-MODIFIED:20260729T211044Z LOCATION:Calit2 URL:https://ics.uci.edu/event/fixed-point-computation-problems-and-facets- of-complexity/ X-SOURCE-NAME:UC Irvine Donald Bren School of Information & Computer Scien ces X-SOURCE-URL:https://ics.uci.edu/event/seattle-area-ics-alumni-reception-2 026 END:VEVENT BEGIN:VEVENT SUMMARY:From Data to Formulas: Statistical Learning for Scalable Symbolic Discovery DTSTART:20261008T230000Z DTEND:20261009T000000Z DTSTAMP:20260826T000618Z UID:10001089-1791475200-1791478800@ics.uci.edu CATEGORIES:education DESCRIPTION:Abstract: Symbolic discovery\, or learning compact mathematica l expressions from data\, offers a form of interpretability in which the f itted model is itself a scientific statement. In this talk\, I treat symbo lic discovery as a statistical learning problem and discuss scalable metho ds for searching large\, structured expression spaces. The starting point is descriptor discovery in materials science\, where candidate predictors are generated from primary physical features through compositions of algeb raic operators\, yielding a combinatorially large and highly correlated pr edictor space. An iterative nonparametric strategy exploits this compositi onal structure to identify\, directly from the primary features\, interpre table descriptors\, or “materials genes\,” associated with binding beh avior in single-atom catalysis. The talk then turns to a key statistical c omponent of scalable symbolic regression: variable selection with Bayesian tree ensembles. Here\, selection accuracy can depend as much on posterior summarization as on the tree prior\, and a simple\, tuning-free posterior summary consistently improves existing approaches and substantially expan ds the reach of symbolic regression. A recent extension moves from regress ion functions to probability laws\, where the discovered expression must s atisfy the structural constraints of a valid distribution. These examples illustrate how statistical formulation through selection\, summarization\, and validity can make interpretable discovery feasible at the scale of mo dern scientific data. LAST-MODIFIED:20260826T000618Z LOCATION:Donald Bren Hall\, Irvine\, CA\, 92697\, United States URL:https://ics.uci.edu/event/from-data-to-formulas-statistical-learning-f or-scalable-symbolic-discovery/ X-SOURCE-NAME:UC Irvine Donald Bren School of Information & Computer Scien ces X-SOURCE-URL:https://ics.uci.edu/event/seattle-area-ics-alumni-reception-2 026 END:VEVENT BEGIN:VEVENT SUMMARY:From Data to Decisions: Reinforcement Learning and Generative AI f or Precision Medicine DTSTART:20261022T230000Z DTEND:20261023T000000Z DTSTAMP:20260821T000606Z UID:10001084-1792684800-1792688400@ics.uci.edu CATEGORIES:education DESCRIPTION:Abstract: Precision medicine aims to transform complex patient data and clinical knowledge into individualized treatment decisions. Howe ver\, this remains challenging when treatments are sequential\, patient st ates evolve over time\, and relevant evidence is distributed across real-w orld data\, clinical guidelines\, and biomedical knowledge. \nIn this talk \, I will discuss how reinforcement learning and generative AI can provide complementary tools for addressing these challenges. Reinforcement learni ng offers a framework for learning adaptive treatment strategies from long itudinal data\, while large language models and knowledge graphs can help organize\, integrate\, and reason over heterogeneous clinical knowledge. I will highlight methodological considerations including incorporation of d omain knowledge\, uncertainty\, treatment coverage\, and interpretability\ , and illustrate these ideas through applications in precision HIV care. LAST-MODIFIED:20260821T000606Z LOCATION:Donald Bren Hall\, Irvine\, CA\, 92697\, United States URL:https://ics.uci.edu/event/from-data-to-decisions-reinforcement-learnin g-and-generative-ai-for-precision-medicine/ X-SOURCE-NAME:UC Irvine Donald Bren School of Information & Computer Scien ces X-SOURCE-URL:https://ics.uci.edu/event/seattle-area-ics-alumni-reception-2 026 END:VEVENT BEGIN:VEVENT SUMMARY:Bayesian Smoothing and Feature Selection via Variational Automatic Relevance Determination DTSTART:20261029T150000Z DTEND:20261030T000000Z DTSTAMP:20260829T000700Z UID:10001090-1793260800-1793293200@ics.uci.edu CATEGORIES:education DESCRIPTION:Abstract: This study introduces Variational Automatic Relevanc e Determination (VARD)\, a novel approach for fitting sparse additive regr ession models in high-dimensional settings. VARD stands out by independent ly assessing the smoothness of each feature while precisely determining wh ether its contribution to the response is zero\, linear\, or nonlinear. Ad ditionally\, we present an efficient coordinate descent algorithm for impl ementing VARD. Empirical evaluations on both simulated and real-world data sets demonstrate VARD’s superior performance compared to alternative var iable selection methods for additive models. \nThis work is in collaborati on with Zihe Liu and Diptarka Saha. LAST-MODIFIED:20260829T000700Z LOCATION:Donald Bren Hall\, Irvine\, CA\, 92697\, United States URL:https://ics.uci.edu/event/bayesian-smoothing-and-feature-selection-via -variational-automatic-relevance-determination-2026/ X-SOURCE-NAME:UC Irvine Donald Bren School of Information & Computer Scien ces X-SOURCE-URL:https://ics.uci.edu/event/seattle-area-ics-alumni-reception-2 026 END:VEVENT BEGIN:VEVENT SUMMARY:Learning-Assisted Inference in High Dimensions: Wasserstein Two-Sa mple Testing DTSTART:20261106T000000Z DTEND:20261106T010000Z DTSTAMP:20260821T000606Z UID:10001085-1793894400-1793898000@ics.uci.edu CATEGORIES:education DESCRIPTION:Abstract: Two-sample testing is a fundamental problem in stati stics\, but detecting general distributional differences becomes increasin gly challenging in high dimensions\, where classical distance-based method s can suffer severely from the curse of dimensionality. In this talk\, I w ill discuss a sequence of approaches that combine optimal transport\, dime nsion reduction\, and statistical learning to address this problem. I will first introduce a test based on the max-sliced Wasserstein distance\, whi ch searches over low-dimensional projections to identify informative direc tions along which two distributions differ\, while permitting statisticall y valid calibration and inference. I will then consider a more flexible le arning-assisted approach in which both projection directions and nonlinear witness functions are learned from the data using optimization and neural networks\, allowing the procedure to adapt to richer forms of distributio nal discrepancy. By combining flexible representation learning with sample splitting and Gaussian approximation\, the resulting tests retain rigorou s statistical guarantees while avoiding prohibitively expensive repeated r esampling. These developments illustrate a broader theme: modern machine-l earning tools can be used to discover informative structure in high-dimens ional data\, while classical statistical principles provide valid and inte rpretable inference. LAST-MODIFIED:20260821T000606Z LOCATION:Donald Bren Hall\, Irvine\, CA\, 92697\, United States URL:https://ics.uci.edu/event/learning-assisted-inference-in-high-dimensio ns-wasserstein-two-sample-testing/ X-SOURCE-NAME:UC Irvine Donald Bren School of Information & Computer Scien ces X-SOURCE-URL:https://ics.uci.edu/event/seattle-area-ics-alumni-reception-2 026 END:VEVENT BEGIN:VEVENT SUMMARY:A Multilayer Network Model for Aggregated Relational Data DTSTART:20261111T000000Z DTEND:20261111T010000Z DTSTAMP:20260821T000606Z UID:10001086-1794326400-1794330000@ics.uci.edu CATEGORIES:education DESCRIPTION:Abstract: The Network Scale-Up Method (NSUM) is a vital tool f or estimating the sizes of hard-to-reach populations using Aggregated Rela tional Data (ARD). Recent advancements in survey design increasingly colle ct ARD across multiple definitions of tie strength\, yielding a complex mu ltilayer network structure. Existing statistical frameworks analyze these layers independently\, sacrificing valuable shared information. We propose a latent space approach for the Multilayer ARD that enables principled jo int estimation. We establish rigorous identifiability conditions for the s hared latent space and mathematically prove that the joint estimator achie ves asymptotic efficiency gains and finite sample bias reductions by “ ”borrowing strength”” across layers. We apply the framework to real- world multilayer NSUM survey data\, demonstrating its practical efficacy i n yielding more robust and precise subpopulation estimates. LAST-MODIFIED:20260821T000606Z LOCATION:Donald Bren Hall\, Irvine\, CA\, 92697\, United States URL:https://ics.uci.edu/event/a-multilayer-network-model-for-aggregated-re lational-data/ X-SOURCE-NAME:UC Irvine Donald Bren School of Information & Computer Scien ces X-SOURCE-URL:https://ics.uci.edu/event/seattle-area-ics-alumni-reception-2 026 END:VEVENT BEGIN:VEVENT SUMMARY:Rating Competitors in Games with Strength-dependent Tie Probabilit ies DTSTART:20261120T000000Z DTEND:20261120T010000Z DTSTAMP:20260821T000606Z UID:10001087-1795104000-1795107600@ics.uci.edu CATEGORIES:education DESCRIPTION:Abstract: Competitor rating systems for head-to-head games are typically used to measure playing strength from game outcomes. Ratings co mputed from these systems are often used to select top competitors for eli te events\, for pairing players of similar strength in online gaming\, and for players to track their own strength over time. Most implemented ratin g systems assume only win/loss outcomes\, and treat occurrences of ties as the equivalent to half a win and half a loss. However\, in games such as chess\, the probability of a tie (draw) is demonstrably higher for stronge r players than for weaker players\, so that rating systems ignoring this a spect of game results may produce strength estimates that are unreliable. We develop a new rating system for head-to-head games that explicitly ackn owledges a tie as a third outcome\, and that the probability of a tie may depend on the strengths of the competitors. Our approach relies on time-va rying game outcomes following a Bayesian dynamic modeling framework\, and that posterior updates within a time period are approximated by one iterat ion of Newton-Raphson evaluated at the prior mean. The approach is demonst rated on a large dataset of chess games played in International Correspond ence Chess Federation tournaments. LAST-MODIFIED:20260821T000606Z LOCATION:Donald Bren Hall\, Irvine\, CA\, 92697\, United States URL:https://ics.uci.edu/event/rating-competitors-in-games-with-strength-de pendent-tie-probabilities/ X-SOURCE-NAME:UC Irvine Donald Bren School of Information & Computer Scien ces X-SOURCE-URL:https://ics.uci.edu/event/seattle-area-ics-alumni-reception-2 026 END:VEVENT BEGIN:VEVENT SUMMARY:Matrix Non-Normal Graphical Model DTSTART:20261204T000000Z DTEND:20261204T010000Z DTSTAMP:20260821T000606Z UID:10001088-1796313600-1796317200@ics.uci.edu CATEGORIES:education DESCRIPTION:Abstract: Contemporary data often have a matrix form\, with in trinsic information embedded in the data structure\, and are high-dimensio nal. The matrix Gaussian graphical model is an important tool for understa nding the dependence structure among such data. A central assumption in th is model is that the observations follow a matrix normal distribution\, wh ich enables characterization of conditional independence via the precision matrix. However\, such an assumption can be too stringent in practice. If data are heavy-tailed\, normality-based methods may produce misleading re sults and unstable estimation. Motivated by this challenge\, we consider t he matrix non-normal graphical model (MANGO). The MANGO model includes the matrix Gaussian graphical model as a special case\, but easily adapts to potential heavy tails. Under this general model\, we investigate the relat ionship between conditional dependence and the precision matrices\, reveal ing that non-normality immediately affects our interpretation of the preci sion matrix. Two sub-models (U-MANGO and H-MANGO) are further studied to a ccommodate different levels of heavy tails. We develop efficient estimatio n procedures for these two models that require minimal additional computat ion over the Gaussian model. Moreover\, two hypothesis tests are developed to identify a suitable model for a particular dataset. Numerical results demonstrate the superior performance of our methods. LAST-MODIFIED:20260821T000606Z LOCATION:Donald Bren Hall\, Irvine\, CA\, 92697\, United States URL:https://ics.uci.edu/event/matrix-non-normal-graphical-model/ X-SOURCE-NAME:UC Irvine Donald Bren School of Information & Computer Scien ces X-SOURCE-URL:https://ics.uci.edu/event/seattle-area-ics-alumni-reception-2 026 END:VEVENT END:VCALENDAR