4:00pm to 5:00pm - https://uci.zoom.us/j/97796361534 - Applied and Computational Mathematics Shuang Li - (UCLA) Non-convex Optimization in Data Science High-dimensional data analysis and estimation appear in many data science and machine learning applications. The underlying low-dimensional structure in these high-dimensional data inspires us to develop optimality guarantees as well as optimization-based techniques for the fundamental problems in data science and machine learning. In recent years, non-convex optimization widely appears in engineering and is solved by many heuristic local algorithms, but lacks global guarantees. The recent geometric/landscape analysis provides a way to determine whether an iterative algorithm can reach global optimality. The landscape of empirical risk has been widely studied in a series of machine learning problems, including low-rank matrix factorization, matrix sensing, matrix completion, and phase retrieval. A favorable geometry guarantees that many algorithms can avoid saddle points and converge to local minima. In this talk, I will introduce some of our recent work on geometric analysis and stochastic algorithms for non-convex optimization problems. |
3:00pm to 3:50pm - RH 306 - Analysis Ming Xiao - (UCSD) K\"ahler-Einstein metrics and obstruction flatness of circle bundles {\bf Abstract:} Obstruction flatness of a strongly pseudoconvex hypersurface $\Sigma$ in a complex manifold refers to the property that any (local) K\"ahler-Einstein metric on the pseudoconvex side of $\Sigma$, complete up to $\Sigma$, has a potential $-\log u$ such that $u$ is $C^\infty$-smooth up to $\Sigma$. In general, $u$ has only a finite degree of smoothness up to $\Sigma$. |
4:00pm to 5:00pm - ISEB 1200 - Differential Geometry Reza Seyyedali - (School of Mathematics, IPM, Iran) Blowup of extremal metrics along submanifolds We give conditions under which the blowup of an extremal Kähler manifold along a submanifold of codimension greater than two admits an extremal metric. This generalizes the work of Arezzo-Pacard-Singer, who considered blowups in points. This is a joint work with Gábor Székelyhidi. |
2:00pm to 3:00pm - 510R Rowland Hall - Combinatorics and Probability Nikita Zhivotovskiy - (UC Berkeley) Estimation of the covariance matrix in the presence of outliers Suppose we are observing a sample of independent random vectors, knowing that the original distribution was contaminated, so that a fraction of observations came from a different distribution. How to estimate the covariance matrix of the original distribution in this case? In this talk, we discuss an estimator of the covariance matrix that achieves the optimal dimension-free rate of convergence under two standard notions of data contamination: We allow the adversary to corrupt a fraction of the sample arbitrarily, while the distribution of the remaining data points only satisfies a certain (rather weak) moment equivalence assumption. Despite requiring the existence of only a few moments, our estimator achieves the same tail estimates as if the underlying distribution were Gaussian. Based on a joint work with Pedro Abdalla. |
9:00am to 10:00am - Zoom - Inverse Problems Graeme Milton - (University of Utah) Untangling in time |
3:00pm to 4:00pm - RH 306 - Number Theory Abhishek Oswal - (Caltech) A p-adic analogue of an algebraization theorem of Borel Let S be a Shimura variety such that the connected components of the set of complex points $S(\mathbb{C})$ are of the form $D/\Gamma$, where $\Gamma$ is a torsion-free arithmetic group acting on the Hermitian symmetric domain $D$. Borel proved that any holomorphic map from any complex algebraic variety into $S(\mathbb{C})$ is an algebraic map. In this talk I shall describe ongoing joint work with Ananth Shankar and Xinwen Zhu, where we prove a $p$-adic analogue of this result of Borel for compact Shimura varieties of abelian type. |
2:00pm to 3:00pm - RH 306 - Nathan Kaplan - (UCI) Codes from varieties over Finite Fields There are $q^{20}$ homogeneous cubic polynomials in four variables with coefficients in the finite field $\mathbb{F}_q$. How many of them define a cubic surface with $q^2+7q+1\ \mathbb{F}_q$-rational points? What about other numbers of rational points? How many of the $q^{20}$ pairs of homogeneous cubic polynomials in three variables define cubic curves that intersect in $9$ $\mathbb{F}_q$-rational points? The goal of this talk is to explain how ideas from the theory of error-correcting codes can be used to study families of varieties over a fixed finite field. We will not assume much background in coding theory and will emphasize examples. |