BEGIN:VCALENDAR
VERSION:2.0
PRODID:researchseminars.org
CALSCALE:GREGORIAN
X-WR-CALNAME:researchseminars.org
BEGIN:VEVENT
SUMMARY:Caroline Uhler (ETH/MIT)
DTSTART:20200605T130000Z
DTEND:20200605T140000Z
DTSTAMP:20260404T110744Z
UID:AlgebraicStatistics/1
DESCRIPTION:Title: Permutations and Posets for Causal Structure Discovery<
/a>\nby Caroline Uhler (ETH/MIT) as part of Algebraic Statistics Online Se
minar\n\nAbstract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/AlgebraicStatistics/1/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Bernd Sturmfels (MPI Leipzig/UC Berkeley)
DTSTART:20200619T190000Z
DTEND:20200619T200000Z
DTSTAMP:20260404T110744Z
UID:AlgebraicStatistics/2
DESCRIPTION:Title: Statistical Models with Rational Maximum Likelihood Est
imator\nby Bernd Sturmfels (MPI Leipzig/UC Berkeley) as part of Algebr
aic Statistics Online Seminar\n\nAbstract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/AlgebraicStatistics/2/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Anna Seigal (University of Oxford)
DTSTART:20200703T130000Z
DTEND:20200703T140000Z
DTSTAMP:20260404T110744Z
UID:AlgebraicStatistics/3
DESCRIPTION:Title: Invariant theory for maximum likelihood estimation\
nby Anna Seigal (University of Oxford) as part of Algebraic Statistics Onl
ine Seminar\n\nAbstract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/AlgebraicStatistics/3/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Ngoc Tran (UT Austin)
DTSTART:20200717T190000Z
DTEND:20200717T200000Z
DTSTAMP:20260404T110744Z
UID:AlgebraicStatistics/4
DESCRIPTION:Title: Graphical models for extreme events with tropical algeb
ra\nby Ngoc Tran (UT Austin) as part of Algebraic Statistics Online Se
minar\n\nAbstract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/AlgebraicStatistics/4/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Seth Sullivant (North Carolina State)
DTSTART:20200828T130000Z
DTEND:20200828T140000Z
DTSTAMP:20260404T110744Z
UID:AlgebraicStatistics/5
DESCRIPTION:Title: Identifiability in phylogenetics using algebraic matroi
ds\nby Seth Sullivant (North Carolina State) as part of Algebraic Stat
istics Online Seminar\n\nAbstract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/AlgebraicStatistics/5/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Joe Kileel (UT Austin)
DTSTART:20200911T190000Z
DTEND:20200911T200000Z
DTSTAMP:20260404T110744Z
UID:AlgebraicStatistics/6
DESCRIPTION:Title: Fast symmetric tensor decomposition\nby Joe Kileel
(UT Austin) as part of Algebraic Statistics Online Seminar\n\nAbstract: TB
A\n
LOCATION:https://stable.researchseminars.org/talk/AlgebraicStatistics/6/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Thomas Kahle (OvGU Magdeburg)
DTSTART:20200925T130000Z
DTEND:20200925T140000Z
DTSTAMP:20260404T110744Z
UID:AlgebraicStatistics/7
DESCRIPTION:Title: Central limit theorems for permutation statistics\n
by Thomas Kahle (OvGU Magdeburg) as part of Algebraic Statistics Online Se
minar\n\nAbstract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/AlgebraicStatistics/7/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Serkan Hosten (San Francisco State University)
DTSTART:20201009T190000Z
DTEND:20201009T200000Z
DTSTAMP:20260404T110744Z
UID:AlgebraicStatistics/8
DESCRIPTION:Title: Two themes on (Gram) spectrahedra: central curves and s
ymmetry\nby Serkan Hosten (San Francisco State University) as part of
Algebraic Statistics Online Seminar\n\n\nAbstract\nThis talk is based on t
wo collaborations: one with Alex Heaton and Isabelle Shankar on symmetry a
dapted Gram spectrahedra\, and the other with Isabelle Shankar and Angelic
a Torres on the degree of the central curve in semidefinite programming (S
DP). The objects in common are spectrahedra. The question for the degree
of the central curve in SDP (where feasible regions are spectrahedra) has
its answer in algebraic statistics as the ML degree of related linear conc
entration models and the relevant geometry of complete quadrics. On the sy
mmetry adapted Gram spectrahedra side\, we use reductions in complexity of
Gram spectrahedra for symmetric polynomials to understand the geometry of
these convex sets. Here I will focus on concrete families and examples.\n
\nZoom link: \nhttps://tum-conf.zoom.us/j/97632429442?pwd=RVNTb3NGb2t4QUdk
SGQzelk5S3luZz09\n
LOCATION:https://stable.researchseminars.org/talk/AlgebraicStatistics/8/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Rina Foygel Barber (University of Chicago)
DTSTART:20201023T180000Z
DTEND:20201023T190000Z
DTSTAMP:20260404T110744Z
UID:AlgebraicStatistics/9
DESCRIPTION:Title: Testing goodness-of-fit and conditional independence wi
th approximate co-sufficient sampling\nby Rina Foygel Barber (Universi
ty of Chicago) as part of Algebraic Statistics Online Seminar\n\n\nAbstrac
t\nGoodness-of-fit (GoF) testing is ubiquitous in statistics\, with direct
ties to model selection\, confidence interval construction\, conditional
independence testing\, and multiple testing\, just to name a few applicati
ons. While testing the GoF of a simple (point) null hypothesis provides an
analyst great flexibility in the choice of test statistic while still ens
uring validity\, most GoF tests for composite null hypotheses are far more
constrained\, as the test statistic must have a tractable distribution ov
er the entire null model space. A notable exception is co-sufficient sampl
ing (CSS): resampling the data conditional on a sufficient statistic for t
he null model guarantees valid GoF testing using any test statistic the an
alyst chooses. But CSS testing requires the null model to have a compact (
in an information-theoretic sense) sufficient statistic\, which only holds
for a very limited class of models\; even for a null model as simple as l
ogistic regression\, CSS testing is powerless. In this paper\, we leverage
the concept of approximate sufficiency to generalize CSS testing to essen
tially any parametric model with an asymptotically-efficient estimator\; w
e call our extension “approximate CSS” (aCSS) testing. We quantify the
finite-sample Type I error inflation of aCSS testing and show that it is
vanishing under standard maximum likelihood asymptotics\, for any choice o
f test statistic. We apply our proposed procedure both theoretically and i
n simulation to a number of models of interest to demonstrate its finite-s
ample Type I error and power. This work is joint with Lucas Janson.\n\nZoo
m link: \nhttps://tum-conf.zoom.us/j/97632429442?pwd=RVNTb3NGb2t4QUdkSGQze
lk5S3luZz09\n
LOCATION:https://stable.researchseminars.org/talk/AlgebraicStatistics/9/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Zehua Lai (University of Chicago)
DTSTART:20201106T200000Z
DTEND:20201106T210000Z
DTSTAMP:20260404T110744Z
UID:AlgebraicStatistics/10
DESCRIPTION:Title: Recht–Re Noncommutative Arithmetic-Geometric Mean Co
njecture is False\nby Zehua Lai (University of Chicago) as part of Alg
ebraic Statistics Online Seminar\n\n\nAbstract\nStochastic optimization al
gorithms have become indispensable in modern machine learning. An importan
t question in this area is the difference between with-replacement samplin
g and without-replacement sampling --- does the latter have superior conve
rgence rate compared to the former? A paper of Recht and Re reduces the pr
oblem to a noncommutative analogue of the arithmetic-geometric mean inequa
lity where n positive numbers are replaced by n positive definite matrices
. If this inequality holds for all n\, then without-replacement sampling (
also known as random reshuffling) indeed outperforms with-replacement samp
ling in some important optimization problems. In this talk\, We will expla
in basic ideas and techniques in polynomial optimization and the theory of
noncommutative Positivstellensatz\, which allows us to reduce the conject
ured inequality to a semidefinite program and the validity of the conjectu
re to certain bounds for the optimum values. Finally\, we show that Recht-
-Re conjecture is false as soon as n=5. We will also discuss some of the c
onsequences of our main theorem. This is a joint work with Lek-Heng Lim.\n
\nZoom link: https://tum-conf.zoom.us/j/97632429442?pwd=RVNTb3NGb2t4QUdkSG
Qzelk5S3luZz09\n
LOCATION:https://stable.researchseminars.org/talk/AlgebraicStatistics/10/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Kaie Kubjas (Aalto University)
DTSTART:20201120T140000Z
DTEND:20201120T150000Z
DTSTAMP:20260404T110744Z
UID:AlgebraicStatistics/11
DESCRIPTION:Title: Uniqueness of nonnegative matrix factorizations\nb
y Kaie Kubjas (Aalto University) as part of Algebraic Statistics Online Se
minar\n\n\nAbstract\nNonnegative matrix factorizations are often encounter
ed in data mining applications where they are used to explain datasets by
a small number of parts. For many of these applications it is desirable th
at there exists a unique nonnegative matrix factorization up to trivial mo
difications given by scalings and permutations. This means that model para
meters are uniquely identifiable from the data. Different sufficient condi
tions for the uniqueness of nonnegative matrix factorizations have been we
ll studied\, however\, a little is known about necessary conditions. We wi
ll give so far the strongest necessary condition for the uniqueness of a n
onnegative factorization. The talk is based on the joint work with Robert
Krone.\n\nZoom link:\n\nhttps://tum-conf.zoom.us/j/97632429442?pwd=RVNTb3N
Gb2t4QUdkSGQzelk5S3luZz09\n
LOCATION:https://stable.researchseminars.org/talk/AlgebraicStatistics/11/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Jose Rodriguez (University of Wisconsin - Madison)
DTSTART:20201204T200000Z
DTEND:20201204T210000Z
DTSTAMP:20260404T110744Z
UID:AlgebraicStatistics/12
DESCRIPTION:Title: Galois Groups in Statistics\nby Jose Rodriguez (Un
iversity of Wisconsin - Madison) as part of Algebraic Statistics Online Se
minar\n\n\nAbstract\nSolving systems of polynomial equations is at the cen
ter of applied algebraic geometry. A common theme of this field is to stud
y a family of systems by allowing some of its coefficients to vary. In alg
ebraic statistics\, the role of coefficients is played by data and the sol
utions we find yield maximum likelihood estimates\, critical points\, and/
or important information about a statistical model. An important invariant
of a family of systems is the Galois (monodromy) group. This captures imp
ortant symmetries within a system and has applications across kinematics\,
computer vision\, power engineering and statistics. In each case\, the Ga
lois group gives a description of how a system's solutions can vary with d
ata.\n\nIn this talk\, I will present three short stories about Galois gro
ups appearing in statistics. The first story emphasizes the idea of treati
ng data as coefficients of a polynomial system. We will visualize the mon
odromy group acting in a nearest point problem where Euclidean distance (E
D) degrees make an appearance. The next story involves Gaussian mixtures a
nd decomposable systems. If time permits\, I will share a third story on h
ow decomposable sparse systems play a role in solving the likelihood equat
ions.\n\nAttendees can join the seminar using the following Zoom link:\n\n
https://tum-conf.zoom.us/j/97632429442?pwd=RVNTb3NGb2t4QUdkSGQzelk5S3luZz0
9\n
LOCATION:https://stable.researchseminars.org/talk/AlgebraicStatistics/12/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Eliana Duarte (OvGU Magdeburg)
DTSTART:20210118T130000Z
DTEND:20210118T140000Z
DTSTAMP:20260404T110744Z
UID:AlgebraicStatistics/13
DESCRIPTION:Title: Algebraic Geometry of Discrete Interventional Models\nby Eliana Duarte (OvGU Magdeburg) as part of Algebraic Statistics Onli
ne Seminar\n\n\nAbstract\nThe Markov equivalence class of a discrete DAG m
odel can be described parametrically via the recursive factorization prope
rty or implicitly by polynomial ideals which are defined via Markov proper
ties of the DAG. We address the problem of describing the Markov equivale
nce classes of discrete DAG models with interventions using polynomial par
ameterizations and vanishing ideals. We show that the algebraic and combin
atorial properties of these models are captured via an interventional stag
ed tree model representation. This point of view leads us to a graphical c
haracterization of the discrete interventional DAG models that are defined
by binomial equations. This is joint work with Liam Solus (KTH\, Sweden)\
, https://arxiv.org/pdf/2012.03593.pdf.\n\nZoom link:\n\nhttps://tum-conf.
zoom.us/j/97632429442?pwd=RVNTb3NGb2t4QUdkSGQzelk5S3luZz09\n
LOCATION:https://stable.researchseminars.org/talk/AlgebraicStatistics/13/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Wenxuan Guo (University of Chicago)
DTSTART:20210201T190000Z
DTEND:20210201T200000Z
DTSTAMP:20260404T110744Z
UID:AlgebraicStatistics/14
DESCRIPTION:Title: Shepp p-product\nby Wenxuan Guo (University of Chi
cago) as part of Algebraic Statistics Online Seminar\n\n\nAbstract\nIn 196
2\, Shepp famously discovered a product of normal random variables that pr
eserves normality. The Shepp product\, which takes the form XY/(X^2 + Y^2)
^1/2\, has since been thoroughly studied and has found numerous connection
s to other areas of statistics. Among other things\, it has an extension t
o n normal variables\, gives a multiplicative analogue of central limit th
eorem\, and applies unexpectedly to genomics as a test statistics for alig
nment-free sequence analysis. The Shepp product is evidently the p = 2 spe
cial case of XY/(X^p + Y^p)^1/p that we call the Shepp p-product. We will
show that the Shepp p-product\, particularly when p = 1 and ∞ (the latte
r in a limiting sense)\, is no less fascinating and applicable than the or
iginal p = 2 case. Just as the Shepp 2-product preserves normal distributi
ons\, the Shepp 1-product preserves Cauchy distributions while the Shepp
∞-product preserves exponential distributions. In fact\, the converse is
also true in an appropriate sense\, allowing us to characterize the Cauch
y\, normal\, and exponential distributions as the unique distributions pre
served by the Shepp p-product for p = 1\, 2\, ∞ respectively. We will st
udy the multiplicative analogue of infinite divisibility with respect to t
he Shepp p-product\, establish an asymptotic theory for the Shepp p-produc
t of n i.i.d. random variables\, and estimate the rates of convergence in
Kolmogorov distance. Alongside our study of convergence rates\, we define
the domain of normal attraction of extremal distributions and establish a
new rate of uniform convergence to Frechet distribution and reverse Weibul
l distribution. Some of our results are new even for the p = 2 case. We wi
ll also discuss new applications of the Shepp p-product in statistics\, co
mputational biology\, and statistical physics. This is joint work with Lek
-Heng Lim.\n\nJoin by Zoom:\n\nhttps://tum-conf.zoom.us/j/97632429442?pwd=
RVNTb3NGb2t4QUdkSGQzelk5S3luZz09\n
LOCATION:https://stable.researchseminars.org/talk/AlgebraicStatistics/14/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Fatemeh Mohammadi (Ghent University)
DTSTART:20210215T130000Z
DTEND:20210215T140000Z
DTSTAMP:20260404T110744Z
UID:AlgebraicStatistics/15
DESCRIPTION:Title: Geometry of conditional independence models with hidde
n variables\nby Fatemeh Mohammadi (Ghent University) as part of Algebr
aic Statistics Online Seminar\n\n\nAbstract\nConditional independence (CI)
is an important tool in statistical modeling\, as\, for example\, it give
s a statistical interpretation to graphical models. In general\, given a l
ist of dependencies among random variables\, it is difficult to say which
constraints are implied by them. Moreover\, it is important to know what c
onstraints on the random variables are caused by hidden variables. On the
other hand\, the CI statements are corresponding to some determinantal con
ditions on the tensor of joint probabilities of the observed random variab
les. Hence\, the geometric analogue of the inference question relates to d
eterminantal varieties and their irreducible decompositions. I will demons
trate how the decompositions of CI varieties lead to interesting algebraic
and combinatorial questions about point configurations in matroid theory
and incidence geometry. This\, in particular\, leads to effective computat
ional approaches for decomposing more general determinantal varieties.\n\n
Zoom link: https://tum-conf.zoom.us/j/97632429442?pwd=RVNTb3NGb2t4QUdkSGQz
elk5S3luZz09\n
LOCATION:https://stable.researchseminars.org/talk/AlgebraicStatistics/15/
END:VEVENT
END:VCALENDAR