BEGIN:VCALENDAR
VERSION:2.0
PRODID:researchseminars.org
CALSCALE:GREGORIAN
X-WR-CALNAME:researchseminars.org
BEGIN:VEVENT
SUMMARY:Chao Wang (UC Davis)
DTSTART:20201013T231000Z
DTEND:20201014T000000Z
DTSTAMP:20260404T094308Z
UID:MADDD_Fall2020/1
DESCRIPTION:Title: From telescope to computed tomography via sparse recovery ap
proaches\nby Chao Wang (UC Davis) as part of Mathematics of Data and D
ecisions at Davis (MADDD) Seminar\n\nAbstract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/MADDD_Fall2020/1/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Cynthia Rudin (Duke)
DTSTART:20201020T231000Z
DTEND:20201021T000000Z
DTSTAMP:20260404T094308Z
UID:MADDD_Fall2020/2
DESCRIPTION:Title: Current Approaches in Interpretable Machine Learning\nby
Cynthia Rudin (Duke) as part of Mathematics of Data and Decisions at Davi
s (MADDD) Seminar\n\nAbstract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/MADDD_Fall2020/2/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Patrice Koehl (UC Davis)
DTSTART:20201027T231000Z
DTEND:20201028T000000Z
DTSTAMP:20260404T094308Z
UID:MADDD_Fall2020/3
DESCRIPTION:Title: Light speed computation of exact solutions to generic and to
degenerate assignment problems\nby Patrice Koehl (UC Davis) as part o
f Mathematics of Data and Decisions at Davis (MADDD) Seminar\n\n\nAbstract
\nThe linear assignment problem is a fundamental problem in combinatorial
optimization with a wide range of applications\, from operational research
to data sciences. It consists of assigning ``agents" to ``tasks" on a one
-to-one basis\, while minimizing the total cost associated with the assign
ment. While many exact algorithms have been developed to identify such an
optimal assignment\, most of these methods are computationally prohibitive
for large size problems. In this talk\, I will describe a novel approach
to solving the assignment problem using techniques adapted from statistica
l physics. In particular I will derive a strongly concave effective free e
nergy function that captures the constraints of the assignment problem at
a finite temperature. This free energy decreases monotonically as a functi
on of $\\beta$\, the inverse of temperature\, to the optimal assignment c
ost\, providing a robust framework for temperature annealing. For large en
ough $\\beta$ values the exact solution to the generic assignment problem
can be derived using a simple round-off to the nearest integer of the elem
ents of the computed assignment matrix. I will also describe a provably co
nvergent method to handle degenerate assignment problems. Finally\, I will
describe computer implementations of this framework that are optimized f
or parallel architectures\, one based on CPU\, the other based on GPU. The
se implementations enable solving large assignment problems (of the orders
of a few 10000s) in computing clock times of the orders of minutes.\n
LOCATION:https://stable.researchseminars.org/talk/MADDD_Fall2020/3/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Patrick Flaherty (UMass)
DTSTART:20201104T001000Z
DTEND:20201104T010000Z
DTSTAMP:20260404T094308Z
UID:MADDD_Fall2020/4
DESCRIPTION:Title: MAP Clustering under the Gaussian Mixture Model via Mixed In
teger Programming\nby Patrick Flaherty (UMass) as part of Mathematics
of Data and Decisions at Davis (MADDD) Seminar\n\n\nAbstract\nIn the appli
cation of clustering models to real data there is often rich prior informa
tion that constrains the relationships among the samples\, or the relation
ships between the samples and the parameters. For example\, in biological
or clinical experiments\, it may be known that two samples are technical r
eplicates and should be assigned to the same cluster\, or it may be known
that the mean value for control samples is in a certain range. However\, s
tandard model-based clustering methods make it difficult to enforce such h
ard logical constraints and may fail to provide a globally optimal cluster
ing. We present a global optimization approach for solving the maximum a-p
osteriori (MAP) clustering problem under the Gaussian mixture model. Our a
pproach can accommodate side constraints and preserves the combinatorial s
tructure of the MAP clustering problem by its formulation as a mixed-integ
er nonlinear optimization problem (MINLP). We approximate the MINLP throug
h a mixed-integer quadratic program (MIQP) transformation that improves co
mputational aspects while guaranteeing $\\epsilon$-global optimality. An i
mportant benefit of our approach is the explicit quantification of the deg
ree of suboptimality\, via the optimality gap\, en route to finding the gl
obally optimal MAP clustering. Numerical experiments comparing our method
to other approaches show that our method finds better optima than standard
clustering methods. Finally\, we cluster a real breast cancer\ngene expre
ssion data set incorporating intrinsic subtype information the induced con
straints substantially improve the computational performance and produce m
ore coherent and biologically meaningful clusters.\n
LOCATION:https://stable.researchseminars.org/talk/MADDD_Fall2020/4/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Zhi Ding (UC Davis)
DTSTART:20201111T001000Z
DTEND:20201111T010000Z
DTSTAMP:20260404T094308Z
UID:MADDD_Fall2020/5
DESCRIPTION:Title: Deep Learning: Not a Simple Hammer for Massive MIMO Wireless
Communication Systems\nby Zhi Ding (UC Davis) as part of Mathematics
of Data and Decisions at Davis (MADDD) Seminar\n\n\nAbstract\nThe prolifer
ation of advanced wireless services\, such as virtual reality\, autonomous
\ndriving and internet of things has generated increasingly intense pressu
re to\ndevelop intelligent wireless communication systems to meet networki
ng needs\nposed by extremely high data rates\, massive number of connected
devices\, and ultra\nlow latency. Deep learning (DL) has been recently em
erged as an exciting design\ntool to advance the development of wireless c
ommunication system with some\ndemonstrated successes. In this talk\, we i
ntroduce the principles of applying DL for\nimproving wireless network per
formance by integrating the underlying\ncharacteristics of channels in pra
ctical massive MIMO deployment. We develop\nimportant insights derived fro
m the physical RF channel properties and present a\ncomprehensive overview
on the application of DL for accurately estimating channel\nstate informa
tion (CSI) of forward channels with low feedback overhead. We\nprovide exa
mples of successful DL application in CSI estimation for massive MIMO\nwir
eless systems and highlight several promising directions for future resear
ch.\n
LOCATION:https://stable.researchseminars.org/talk/MADDD_Fall2020/5/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Chelsea Weaver (Amazon)
DTSTART:20201118T001000Z
DTEND:20201118T010000Z
DTSTAMP:20260404T094308Z
UID:MADDD_Fall2020/6
DESCRIPTION:Title: Natural Language Understanding at Amazon Music\nby Chels
ea Weaver (Amazon) as part of Mathematics of Data and Decisions at Davis (
MADDD) Seminar\n\n\nAbstract\nIn this talk\, I’ll discuss what happens w
hen you ask an Alexa device to play music. I’ll focus on the Natural Lan
guage Understanding (NLU) component\, which deals with categorizing and la
beling transcribed requests. In particular\, I’ll discuss two projects I
’ve worked on designed to improve upon the initial labeling. The first u
ses a BERT-based language model to “correct” requests that appear to b
e mislabeled. The second is an online learning model that selects from dif
ferent NLU interpretations using implicit customer feedback. I’ll conclu
de the talk with a few tips for the industry job search.\n\nLinks: Amazon
Jobs Page\; Amazon Science Page\n\nPapers:\nPersonalizing natural-language
understanding using multi-armed bandits and implicit feedback – Moerche
n et al (2020)\nCounterfactual Risk Minimization: Learning from Logged Ban
dit Feedback - Swaminathan & Joachims (2015)\nAnalysis of Thompson Sampli
ng for the Multi-Armed Bandit Problem – Agrawal et al (2012)\n
LOCATION:https://stable.researchseminars.org/talk/MADDD_Fall2020/6/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Wolfgang Polonik (UC Davis)
DTSTART:20201125T001000Z
DTEND:20201125T010000Z
DTSTAMP:20260404T094308Z
UID:MADDD_Fall2020/7
DESCRIPTION:Title: Multiscale Geometric Feature Extraction\nby Wolfgang Pol
onik (UC Davis) as part of Mathematics of Data and Decisions at Davis (MAD
DD) Seminar\n\n\nAbstract\nA method for extracting multiscale geometric fe
atures from a data cloud is presented. Each pair of data points is mapped
into a real-valued feature function\, whose construction is based on geome
tric considerations. The collection of these feature functions is then bei
ng used for further data analysis. Applications include classification\, a
nomaly detection and data visualization. In contrast to the popular kernel
trick\, the construction of the feature functions is based on geometric c
onsiderations. The performance of the methodology is illustrated through a
pplications to real data sets\, and some theoretical guarantees supporting
the performance of the novel methodology are presented. This is joint wor
k with G. Chandler.\n
LOCATION:https://stable.researchseminars.org/talk/MADDD_Fall2020/7/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Guido Montufar (UCLA)
DTSTART:20201202T001000Z
DTEND:20201202T010000Z
DTSTAMP:20260404T094308Z
UID:MADDD_Fall2020/8
DESCRIPTION:Title: Optimal Transport to Independence Models\nby Guido Montu
far (UCLA) as part of Mathematics of Data and Decisions at Davis (MADDD) S
eminar\n\n\nAbstract\nAn independence model for discrete random variables
is a Segre-Veronese variety in a probability simplex. Any metric on the se
t of joint states of the random variables induces a Wasserstein metric on
the probability simplex. The unit ball of this polyhedral norm is dual to
the Lipschitz polytope. Given any data distribution\, we seek to minimize
its Wasserstein distance to a fixed independence model. The solution to th
is optimization problem is a piecewise algebraic function of the data. We
compute this function explicitly in small instances\, we examine its combi
natorial structure and algebraic degrees in the general case\, and we pres
ent some experimental case studies. This talk is based on joint work with
Türkü Özlüm Çelik\, Asgar Jamneshan\, Bernd Sturmfels\, Lorenzo Ventu
rello.\n\nhttps://arxiv.org/abs/1909.11716\n\nhttps://arxiv.org/abs/2003.0
6725\n
LOCATION:https://stable.researchseminars.org/talk/MADDD_Fall2020/8/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Samir Chowdhury (Stanford)
DTSTART:20201209T001000Z
DTEND:20201209T010000Z
DTSTAMP:20260404T094308Z
UID:MADDD_Fall2020/9
DESCRIPTION:Title: Gromov-Wasserstein Learning in a Riemannian Framework\nb
y Samir Chowdhury (Stanford) as part of Mathematics of Data and Decisions
at Davis (MADDD) Seminar\n\n\nAbstract\nGeometric and topological data ana
lysis methods are increasingly being used to derive insights from data ari
sing in the empirical sciences. We start with a particular case where such
techniques are applied to human neuroimaging data to obtain graphs which
can then yield insights connecting neurobiology to human task performance.
Reproducing such insights across populations requires statistical learnin
g techniques such as averaging and PCA across graphs without known node co
rrespondences. We formulate this problem using the Gromov-Wasserstein (GW)
distance and present a recently-developed Riemannian framework for GW-ave
raging and tangent PCA. Beyond graph adjacency matrices\, this framework p
ermits consuming derived network representations such as distance or kerne
l matrices\, and each choice leads to additional structure on the GW probl
em that can be exploited for theoretical and/or computational advantages.
In particular\, we show how replacing the adjacency matrix representation
with a spectral representation leads to theoretical guarantees allowing ef
ficient use of the Riemannian framework. Additionally we present numerics
showing how the spectral representation achieves state of the art accuracy
and runtime in graph learning tasks such as matching and partitioning on
a variety of real and simulated datasets.\n
LOCATION:https://stable.researchseminars.org/talk/MADDD_Fall2020/9/
END:VEVENT
END:VCALENDAR