BEGIN:VCALENDAR
VERSION:2.0
PRODID:researchseminars.org
CALSCALE:GREGORIAN
X-WR-CALNAME:researchseminars.org
BEGIN:VEVENT
SUMMARY:Victor Barranca (Swarthmore College)
DTSTART;VALUE=DATE-TIME:20221129T171500Z
DTEND;VALUE=DATE-TIME:20221129T181500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/1
DESCRIPTION:Title: Reconstruction of Neuronal Network Connectivity and Rival
rous Percepts Via Compressive Sensing of Network Dynamics\nby Victor B
arranca (Swarthmore College) as part of Northwestern Applied Mathematics S
eminar\n\nLecture held in M416 Tech Northwestern Evanston.\n\nAbstract\nNe
uronal network connectivity demonstrates sparsity on multiple spatial scal
es and natural stimuli also possess sparse representations in numerous dom
ains. In this talk\, we underline the role of sparsity in the efficient en
coding of network connectivity and inputs through nonlinear neuronal netwo
rk dynamics. Addressing the fundamental challenge of recovering the struct
ural connectivity of large-scale neuronal networks\, we leverage propertie
s of the balanced dynamical regime and compressive sensing theory to devel
op a theoretical framework for efficiently reconstructing sparse network c
onnections through measurements of the network response to a relatively sm
all ensemble of random stimuli. We further utilize sparse recovery ideas t
o probe the neural correlates of binocular rivalry through dynamic percept
reconstructions based on the activity of a two-layer network model with c
ompeting downstream pools driven by disparate image stimuli. The resultant
model dynamics agree with key experimental observations and give insights
into the excitation/inhibition hypothesis for autism.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/1/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Richard Braun (University of Delaware)
DTSTART;VALUE=DATE-TIME:20230131T171500Z
DTEND;VALUE=DATE-TIME:20230131T181500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/2
DESCRIPTION:Title: Semi-automated Tear Breakup Detection and Modeling on the
Ocular Surface\nby Richard Braun (University of Delaware) as part of
Northwestern Applied Mathematics Seminar\n\nLecture held in M416 Tech Nort
hwestern Evanston.\n\nAbstract\nThe tear film is a thin fluid multilayer l
eft on the eye surface after a blink. A good tear film is essential for h
ealth and proper function of the eye. Millions of people have a condition
called dry eye disease (DED) that is thought to be closely linked to the
tear film. DED inhibits vision and may lead to inflammation and ocular su
rface damage. However\, there is little quantitative data about tear film
failure\, often called tear break up (TBU). Currently\, it is not possibl
e to directly measure important variables such as tear osmolarity (saltine
ss) within areas of TBU. We present a mostly automatic method that we have
developed to extract data from video of the tear film dyed with fluoresce
in (for visualization). We have extracted data for 15 healthy subjects res
ulting in 467 instances of TBU. Using parameter identification from fits t
o appropriate math models\, we estimate which mechanisms are most importan
t in each instance and determine critical variables such as osmolarity wit
hin regions of TBU. Not only is new data obtained\, but far more data\, en
abling statistical methods to be applied. So far\, the methods provide bas
eline data for TBU in healthy subjects\; future work will produce data fro
m DED subjects.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/2/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Keaton Burns (Massachusettes Institute of Technology)
DTSTART;VALUE=DATE-TIME:20230207T171500Z
DTEND;VALUE=DATE-TIME:20230207T181500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/3
DESCRIPTION:Title: Solving Partial Differential Equations Exactly Over Polyn
omials\nby Keaton Burns (Massachusettes Institute of Technology) as pa
rt of Northwestern Applied Mathematics Seminar\n\nLecture held in M416 Tec
h Northwestern Evanston.\nAbstract: TBA\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/3/
END:VEVENT
BEGIN:VEVENT
SUMMARY:David Bortz (University of Colorado Boulder)
DTSTART;VALUE=DATE-TIME:20230221T171500Z
DTEND;VALUE=DATE-TIME:20230221T181500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/5
DESCRIPTION:Title: The Surprising Robustness and Computational Efficiency of
Weak Form System Identification\nby David Bortz (University of Colora
do Boulder) as part of Northwestern Applied Mathematics Seminar\n\nLecture
held in M416 Tech Northwestern Evanston.\nAbstract: TBA\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/5/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Daniel Wells (Santa Ana Bio\, Inc.)
DTSTART;VALUE=DATE-TIME:20230228T171500Z
DTEND;VALUE=DATE-TIME:20230228T181500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/7
DESCRIPTION:Title: Single Cell Spatial Transcriptomics to Accelerate Systems
Immunology\nby Daniel Wells (Santa Ana Bio\, Inc.) as part of Northwe
stern Applied Mathematics Seminar\n\nLecture held in M416 Tech Northwester
n Evanston.\n\nAbstract\nSingle cell transcriptomics has revolutionized im
munology and driven the emergence of systems-level approaches in immunolog
y. However\, existing technologies work only on dissociated cells and lose
valuable tissue context. Single cell spatial transcriptomics is an emergi
ng field which leverages large scale profiling of RNA and protein in situ
to measure the state of individual cells within intact tissues. Here we wi
ll provide an overview of single cell genomic approaches\, their uses\, an
d demonstrate ways single cell spatial transcriptomics can augment underst
anding of immunology. Particular focus will be paid to emerging analytic a
pproaches to extract differentiated signal from these data. FInally\, we w
ill provide an example of how these approaches can be applied in the setti
ng of autoimmunity.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/7/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Eva Kanso (University of Southern California)
DTSTART;VALUE=DATE-TIME:20230404T161500Z
DTEND;VALUE=DATE-TIME:20230404T171500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/8
DESCRIPTION:by Eva Kanso (University of Southern California) as part of No
rthwestern Applied Mathematics Seminar\n\nLecture held in M416 Tech Northw
estern Evanston IL.\nAbstract: TBA\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/8/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Simone Bianco (Altos Labs - Bay Area Institute of Science)
DTSTART;VALUE=DATE-TIME:20230411T161500Z
DTEND;VALUE=DATE-TIME:20230411T171500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/9
DESCRIPTION:by Simone Bianco (Altos Labs - Bay Area Institute of Science)
as part of Northwestern Applied Mathematics Seminar\n\nLecture held in M41
6 Tech Northwestern Evanston IL.\nAbstract: TBA\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/9/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Gwynn Elfring (University of British Columbia)
DTSTART;VALUE=DATE-TIME:20230418T161500Z
DTEND;VALUE=DATE-TIME:20230418T171500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/10
DESCRIPTION:Title: Active Matter in Inhomogeneous Environments\nby Gwyn
n Elfring (University of British Columbia) as part of Northwestern Applied
Mathematics Seminar\n\nLecture held in M416 Tech Northwestern Evanston IL
.\nAbstract: TBA\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/10/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Osman Basaran (Purdue University)
DTSTART;VALUE=DATE-TIME:20230502T161500Z
DTEND;VALUE=DATE-TIME:20230502T171500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/11
DESCRIPTION:by Osman Basaran (Purdue University) as part of Northwestern A
pplied Mathematics Seminar\n\nLecture held in M416 Tech Northwestern Evans
ton IL.\nAbstract: TBA\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/11/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Sung Ha Kang (Georgia Institute of Technology)
DTSTART;VALUE=DATE-TIME:20230509T161500Z
DTEND;VALUE=DATE-TIME:20230509T171500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/12
DESCRIPTION:Title: Identifying Differential Equations with Numerical Method
s: Time Evolution\, Subspace Pursuit and Weak Form\nby Sung Ha Kang (G
eorgia Institute of Technology) as part of Northwestern Applied Mathematic
s Seminar\n\nLecture held in M416 Tech Northwestern Evanston IL.\nAbstract
: TBA\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/12/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Tiffany Shaw (University of Chicago)
DTSTART;VALUE=DATE-TIME:20230516T161500Z
DTEND;VALUE=DATE-TIME:20230516T171500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/13
DESCRIPTION:Title: Fast jet stream winds get faster under climate change\nby Tiffany Shaw (University of Chicago) as part of Northwestern Applied
Mathematics Seminar\n\nLecture held in M416 Tech Northwestern Evanston IL
.\nAbstract: TBA\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/13/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Linda Petzold (University of California\, Santa Barbara)
DTSTART;VALUE=DATE-TIME:20230522T210000Z
DTEND;VALUE=DATE-TIME:20230522T221500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/14
DESCRIPTION:Title: The Roles and Consequences of Randomness in Biological S
ystems\nby Linda Petzold (University of California\, Santa Barbara) as
part of Northwestern Applied Mathematics Seminar\n\nLecture held in M416
Tech Northwestern Evanston IL.\nAbstract: TBA\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/14/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Linda Petzold (University of California\, Santa Barbara)
DTSTART;VALUE=DATE-TIME:20230523T161500Z
DTEND;VALUE=DATE-TIME:20230523T171500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/15
DESCRIPTION:Title: Interpretable Polynomial Neural ODEs\nby Linda Petzo
ld (University of California\, Santa Barbara) as part of Northwestern Appl
ied Mathematics Seminar\n\nLecture held in M416 Tech Northwestern Evanston
IL.\nAbstract: TBA\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/15/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Christine Heitsch (Georgia Institute of Technology)
DTSTART;VALUE=DATE-TIME:20231003T161500Z
DTEND;VALUE=DATE-TIME:20231003T171500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/17
DESCRIPTION:Title: How Can Discrete Mathematics Improve RNA Folding Predict
ions\nby Christine Heitsch (Georgia Institute of Technology) as part o
f Northwestern Applied Mathematics Seminar\n\nLecture held in M416 Tech No
rthwestern Evanston IL.\nAbstract: TBA\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/17/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Aleksandra Walczak (ENS\, Paris)
DTSTART;VALUE=DATE-TIME:20231010T161500Z
DTEND;VALUE=DATE-TIME:20231010T171500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/18
DESCRIPTION:Title: Mathematics of Life Series: Formation of Immune Repertoi
re\nby Aleksandra Walczak (ENS\, Paris) as part of Northwestern Applie
d Mathematics Seminar\n\nLecture held in M416 Tech Northwestern Evanston I
L.\n\nAbstract\nLink: https://northwestern.zoom.us/j/94392051105\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/18/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Sam Kriegman (Northwestern University)
DTSTART;VALUE=DATE-TIME:20231017T161500Z
DTEND;VALUE=DATE-TIME:20231017T171500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/19
DESCRIPTION:Title: Teaching Evolution Calculus: Efficient Automatic Design
of Robots\nby Sam Kriegman (Northwestern University) as part of Northw
estern Applied Mathematics Seminar\n\nLecture held in M416 Tech Northweste
rn Evanston IL.\nAbstract: TBA\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/19/
END:VEVENT
BEGIN:VEVENT
SUMMARY:M. Graham (UW Madison)
DTSTART;VALUE=DATE-TIME:20231023T210000Z
DTEND;VALUE=DATE-TIME:20231023T220000Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/20
DESCRIPTION:Title: Data\, Dynamics\, and Manifolds: Machine Learning Approa
ches for Modeling and Controlling Complex Flows\nby M. Graham (UW Madi
son) as part of Northwestern Applied Mathematics Seminar\n\nLecture held i
n M416 Tech Northwestern Evanston IL.\nAbstract: TBA\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/20/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Sky Nicholson (Northwestern University)
DTSTART;VALUE=DATE-TIME:20231031T161500Z
DTEND;VALUE=DATE-TIME:20231031T171500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/21
DESCRIPTION:Title: How to Quantify Rare-Events From Microscopic Kinetics Us
ing Tensor Networks\nby Sky Nicholson (Northwestern University) as par
t of Northwestern Applied Mathematics Seminar\n\nLecture held in M416 Tech
Northwestern Evanston IL.\n\nAbstract\nMolecules can undergo reactions an
d diffusion through space\, creating the cornucopia of patterns we observe
in nature. Understanding how these patterns emerge is challenging to stud
y due to the immense separation of scales between the fast microscopic dyn
amics and the macroscopic pattern. A classic example of such a pattern is
bistability\, where a system will spontaneously switch between two macrosc
opic states of the system. Quantifying the rate of switching historically
has relied on waiting for exponentially rare events in the system to be ob
served. Ensembles of such events lead to estimates of kinetic rates. In th
is work we show how to calculate rare macroscopic rates from high-dimensio
nal reaction diffusion systems without resorting to sampling techniques. I
nstead\, we exploit and extract observables such as macroscopic rates by e
volving the ensemble of all possible trajectories. The foundation of this
work is based on using the Doi-Peliti formalism to encode the chemical mas
ter equation into a second-quantized form. This form allows chemical netwo
rks to be readily evolved using efficient tensor network methods. Our resu
lts are illustrated using an adapted version of the bistable Sch¨ogl mode
l with diffusion. We calculate rates over five-orders of magnitude for lar
ge systems (∼ 3 × 1015 microstates) and show strong agreement to kineti
c-Monto Carlo simulations and the more advanced forward flux sampling meth
od. Our Doi-Peliti tensor network procedure demonstrates sub-exponential s
caling in computational expense\, while bypassing complications due to sam
pling errors or needing intimate knowledge of the reaction network as is t
he case with more advanced sampling methods.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/21/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Thierry Mora (ENS\, Paris)
DTSTART;VALUE=DATE-TIME:20231107T171500Z
DTEND;VALUE=DATE-TIME:20231107T181500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/22
DESCRIPTION:Title: Mathematics of Life Series: Statistical Mechanics of Col
lective Behavior\nby Thierry Mora (ENS\, Paris) as part of Northwester
n Applied Mathematics Seminar\n\nLecture held in M416 Tech Northwestern Ev
anston IL.\n\nAbstract\nSome animal groups behave in a highly coordinated
way\, reminiscent of ordered phases in physics. However\, animals are also
heterogeneous\, have memory\, and operate out of equilibrium. I will pres
ent recent attempts at modeling the complex dynamics of social groups of m
ice interacting freely in a controlled environment. I will then assess how
far from equilibrium collective behaviour might be\, both in recordings o
f real bird flocks and in flocking models.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/22/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Milo Lin (UT Southwestern)
DTSTART;VALUE=DATE-TIME:20231128T171500Z
DTEND;VALUE=DATE-TIME:20231128T181500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/23
DESCRIPTION:Title: Mathematics of Life Seminar Series: Thermodynamic Limits
of Molecular Computation\nby Milo Lin (UT Southwestern) as part of No
rthwestern Applied Mathematics Seminar\n\nLecture held in M416 Tech Northw
estern Evanston IL.\n\nAbstract\nLiving systems update their status by alt
ering the probability distribution of stochastically rearranging molecules
in response to a change in the system parameters. These updates constitut
e molecular computational steps. Due to the presence of thermodynamic driv
ing forces\, typically in the form of chemical gradients\, these computati
ons convert a molecular system from one non-equilibrium steady state to an
other. Because such steady states are energetically costly to maintain\, t
he question arises as to why nature has evolved this computational scheme.
I will discuss a thermodynamic limit on computation. Namely\, for any mol
ecular system performing any computational step\, the maximum information
gained in the computation is shown to be a simple function of the thermody
namic force. Therefore\, the presence of thermodynamic forces\, and the ex
penditure of energy\, allows biomolecular systems to convert modest change
s in input into striking changes in output that would be surprising or imp
ossible at equilibrium.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/23/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Niall Mangan and Katelyn Leisman (Northwestern University)
DTSTART;VALUE=DATE-TIME:20231114T171500Z
DTEND;VALUE=DATE-TIME:20231114T181500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/24
DESCRIPTION:Title: Flushed with Insights: The Promising Potential of Poop-B
ased Testing for Public Health\nby Niall Mangan and Katelyn Leisman (N
orthwestern University) as part of Northwestern Applied Mathematics Semina
r\n\nLecture held in M416 Tech Northwestern Evanston IL.\n\nAbstract\nEsti
mating the prevalence of infectious disease in a community is useful for p
ublic health resource allocation\, policy making\, and messaging. When dis
eases such as COVID-19 become endemic in the community it is essential to
have passive indicators that do not depend on voluntary testing data. Our
team is working with public health departments to use wastewater to inform
our understanding of COVID-19 prevalence in communities throughout Illino
is. We have developed a generalized methodology to improve the predictive
power of wastewater from treatment plants in the Chicago area. Connecting
measured SARS-CoV-2 RNA to community prevalence is challenging\, due to ch
anges in the contributing population\, the variable rate of wastewater flo
w\, and the complexity of wastewater media\, which impacts RNA decay rates
and lab measurement accuracy. To quantify the impact of these factors we
also track other viruses including pepper mild mottle virus (PMMoV)\, a bi
omarker for the number of people contributing to the wastewater\, and bovi
ne coronavirus (BCoV)\, a lab process recovery control. We build and compa
re a set of multi-linear regression models\, which incorporate PMMoV\, BCo
V\, and flow rate into a corrected estimate for SARS-CoV-2 RNA concentrati
on. Laboratory methods evolved rapidly during the COVID-19 pandemic\, and
we show that correction terms differ depending on the laboratory procedure
s used in analyzing the samples. Nonetheless\, in all cases a statistical
correction model provides a significant improvement in terms of correlatio
n with hospitalizations and trend analysis over doing no correction.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/24/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Albane Thery (University of Pennsylvania)
DTSTART;VALUE=DATE-TIME:20240123T171500Z
DTEND;VALUE=DATE-TIME:20240123T181500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/26
DESCRIPTION:Title: Building Models For Swimmers in Complex and Confined Env
ironments\nby Albane Thery (University of Pennsylvania) as part of Nor
thwestern Applied Mathematics Seminar\n\nLecture held in M416 Tech Northwe
stern Evanston IL.\nAbstract: TBA\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/26/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Yue Yu (Lehigh University)
DTSTART;VALUE=DATE-TIME:20240206T171500Z
DTEND;VALUE=DATE-TIME:20240206T181500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/27
DESCRIPTION:Title: Nonlocal Operator is All You Need\nby Yue Yu (Lehigh
University) as part of Northwestern Applied Mathematics Seminar\n\nLectur
e held in M416 Tech Northwestern Evanston IL.\n\nAbstract\nDuring the last
20 years there has been a lot of progress in applying neural networks (NN
s) to many machine learning tasks. However\, their employment in scientifi
c machine learning with the purpose of learning physics of complex system
is less explored. Differs from the other machine learning tasks such as th
e computer vision and natural language processing problems where a large a
mount of unstructured data are available\, physics-based machine learning
tasks often feature scarce and structured measurements. \nIn this talk
\, we will take the learning of heterogeneous material responses as an exe
mplar problem\, to investigate the design of neural networks for physics-b
ased machine learning. In particular\, we propose to parameterize the mapp
ing between loading conditions and the corresponding system responses in t
he form of nonlocal neural operators\, and infer the neural network parame
ters from high-fidelity simulation or experimental measurements. As such\,
the model is built as mappings between infinite-dimensional function spac
es\, and the learnt network parameters are resolution-agnostic: no further
modification or tuning will be required for different resolutions in orde
r to achieve the same level of prediction accuracy. Moreover\, the nonloca
l operator architecture also allows the incorporation of intrinsic mathema
tical and physics knowledge\, which improves the learning efficacy and rob
ustness from scarce measurements. \nTo demonstrate the applicability o
f our nonlocal operator learning framework\, three typical scenarios in ph
ysics-based machine learning will be discussed: the learning of a material
-specific constitutive law\, the learning of an efficient PDE solution ope
rator\, and the development of a foundational constitutive law across mult
iple materials. As an application\, we learn material models directly from
digital image correlation (DIC) displacement tracking measurements on a p
orcine tricuspid valve leaflet tissue\, and show that the learnt model sub
stantially outperforms conventional constitutive models.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/27/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Geoff Vallis (Exeter)
DTSTART;VALUE=DATE-TIME:20240213T171500Z
DTEND;VALUE=DATE-TIME:20240213T181500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/28
DESCRIPTION:Title: Rainy-Bénard Convection: An Idealization of a Moist Atm
osphere\nby Geoff Vallis (Exeter) as part of Northwestern Applied Math
ematics Seminar\n\nLecture held in M416 Tech Northwestern Evanston IL.\n\n
Abstract\nRayleigh-Benard convection is commonly regarded as the benchmark
system for convection in a wide variety of settings\, and its simplicity
has led to a great many theoretical\, experimental and computational studi
es. However\, it is often regarded as irrelevant as a model for convection
in Earth’s atmosphere because of the significant\, almost dominating\,
influence of moisture in the latter system\, in addition to many other com
plications. In an attempt to partially bridge the evident gap between thes
e systems we add a condensate to the Rayleigh-Benard system but keep other
aspects the same. The resulting ‘Rainy-Benard’ system has very rich b
ehavior and in this talk I’ll describe some that behavior and other prop
erties of the system.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/28/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Sid Goyal (University of Toronto)
DTSTART;VALUE=DATE-TIME:20240220T171500Z
DTEND;VALUE=DATE-TIME:20240220T181500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/29
DESCRIPTION:Title: POSTPONED Mathematics of Life Seminar Series: Competitio
n Across Scales in Biology\nby Sid Goyal (University of Toronto) as pa
rt of Northwestern Applied Mathematics Seminar\n\nLecture held in M416 Tec
h Northwestern Evanston IL.\nAbstract: TBA\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/29/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Snezhana Abarzhi (U Western Adelaide (Australia))
DTSTART;VALUE=DATE-TIME:20240227T171500Z
DTEND;VALUE=DATE-TIME:20240227T181500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/30
DESCRIPTION:Title: Interface Dynamics in Ideal and Realistic Fluids\nby
Snezhana Abarzhi (U Western Adelaide (Australia)) as part of Northwestern
Applied Mathematics Seminar\n\nLecture held in M416 Tech Northwestern Eva
nston IL.\n\nAbstract\nInterface and mixing and their non-equilibrium kine
tics and dynamics couple micro to macro scales\, and are ubiquitous to occ
ur in fluids\, plasmas and materials. Stellar evolution\, plasma fusion\,
reactive fluids\, microfluidics\, purification of water\, and nanofabricat
ion are a few examples of many processes to which dynamics of interfaces i
s directly relevant. This talk presents the rigorous theory of the stabili
ty of the interface – a phase boundary broadly defined. We directly link
the structure of macroscopic flow fields to microscopic interfacial trans
port\, quantify the contributions of macro and micro stabilization mechani
sms to interface stability\, and discover the fluid instabilities never pr
eviously discussed. In ideal and realistic fluids\, the interface stabilit
y is set primarily by the interplay of the macroscopic inertial mechanism
balancing the destabilizing acceleration\, whereas microscopic thermodynam
ics create vortical fields in the bulk. By linking micro to macro scales\,
the interface is the place where balances are achieved.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/30/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Peko Hosoi (Massachusetts Institute of Technology (MIT))
DTSTART;VALUE=DATE-TIME:20240326T161500Z
DTEND;VALUE=DATE-TIME:20240326T171500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/31
DESCRIPTION:Title: Filtration and Fluid Mechanics Inspired by the Manta Ray
- Reiss Lecture\nby Peko Hosoi (Massachusetts Institute of Technology
(MIT)) as part of Northwestern Applied Mathematics Seminar\n\nLecture hel
d in M416 Tech Northwestern Evanston IL.\n\nAbstract\nHVAC systems account
for about 20% of U.S. energy consumption of which at least 7% is consumed
by fans. Their energy efficiency strongly depends on their filters: reduc
ing resistance can result in significant energy savings. We explore novel
strategies for filtration inspired by the manta ray\, which has evolved a
system for filtering zooplankton that appears to be unlike any industrial
filtration mechanism. Instead of a sieve strategy\, the manta deploys micr
ostructures\, which are hypothesized to instigate eddies that push particl
es away from the filtration pores\, resisting clogging\, and enabling the
filtration of particles much smaller than the pore size. Using perturbatio
n theory and asymptotics we examine two toy problems that mimic various fe
atures of the filtration strategies employed by manta rays and find that e
xperimental data from wavy channels are consistent with our asymptotic pre
dictions.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/31/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Efi Efrati (Weizmann Institute of Science)
DTSTART;VALUE=DATE-TIME:20240416T161500Z
DTEND;VALUE=DATE-TIME:20240416T171500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/32
DESCRIPTION:Title: Frustrated Assemblies: Describing Matter from Within
\nby Efi Efrati (Weizmann Institute of Science) as part of Northwestern Ap
plied Mathematics Seminar\n\nLecture held in M416 Tech Northwestern Evanst
on IL.\n\nAbstract\nFrustrated assemblies are comprised of ill-fitting bui
lding blocks whose favored local relative arrangement cannot be globally r
ealized. In fact\, most self assembled structures contain some degree of f
rustration. In some cases\, the frustration is locally resolved and leads
to little or no structural consequence\, while in other cases it dominates
the assembly's size\, shape\, and response properties. In this talk\,
I will present different manifestations of geometric frustration as they
naturally arise in molecular crystals\, liquid crystals\, spin systems\, a
nd nanoparticle assemblies. To theoretically study these assemblies we res
ort to an intrinsic approach in which matter is described only through loc
al properties available to an observer residing within the material. The f
rustration is then quantified by the geometric compatibility conditions wh
ose structure allows us to classify the frustration and predict the assemb
ly's behavior without explicitly solving for the ground state of the syste
m.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/32/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Ehud Yariv (Technion - Israel Institute of Technology)
DTSTART;VALUE=DATE-TIME:20240430T161500Z
DTEND;VALUE=DATE-TIME:20240430T171500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/33
DESCRIPTION:Title: Shocks and Caps in Drop Electrohydrodynamics\nby Ehu
d Yariv (Technion - Israel Institute of Technology) as part of Northwester
n Applied Mathematics Seminar\n\nLecture held in M416 Tech Northwestern Ev
anston IL.\n\nAbstract\nThe problem of electrohydrodynamic drop deformatio
n is well understood in the case where the external electric field is weak
. In one of his many celebrated papers (Proc. R. Soc. A\, 291 1425 159-166
\, 1966)\, G. I. Taylor worked out a complete theory in this limit\, inclu
ding analytical expressions for the electrohydrodynamic flow engendered wi
thin and outside of the drop by the electric field acting on its own induc
ed interfacial charge\, and a simple function of the permittivity\, conduc
tivity and viscosity drop-to-background ratios discriminating between prol
ate or oblate deformation. In this talk\, we will employ numerical a
nd asymptotic tools to explore the effects of interfacial-charge convectio
n\, which were neglected by Taylor but become important at strong electric
fields. In particular\, we will analyze (in 2D\, for simplicity) how Tayl
or’s fore-aft symmetric solution evolves as the electrical Reynolds numb
er is increased from zero to arbitrarily large values. What we shall find
is hinted by the title of the talk. This is joint work with Gunnar G. Peng
\, Rodolfo Brandão and Ory Schnitzer.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/33/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Pankaj Mehta (Boston University)
DTSTART;VALUE=DATE-TIME:20240507T161500Z
DTEND;VALUE=DATE-TIME:20240507T171500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/34
DESCRIPTION:Title: Mathematics of Life Seminar Series: Randomness\, Complex
ity\, and the Biological Frontier\nby Pankaj Mehta (Boston University)
as part of Northwestern Applied Mathematics Seminar\n\nLecture held in M4
16 Tech Northwestern Evanston IL.\n\nAbstract\nThe towering successes of t
wentieth century theoretical physics were marked by two guiding principles
: symmetry and energy functionals (reflecting equilibrium dynamics). Yet h
ow we can exploit these principles to develop a theory of living systems i
s unclear since the biological world is composed of heterogeneous\, intera
cting components operating out of equilibrium. In this talk\, I will argue
that one possible strategy for taming biological complexity is to embrace
the idea that many biological behaviors we observe are “typical” and
can be modeled using random systems that respect biologically-inspired con
straints. I will focus on showing how this approach can be used to make cl
ose connection with experiments by presenting three vignettes focusing on:
(i) theory-inspired techniques for visualizing single-cell transcriptomic
s data for cellular identity\, (ii) understanding how the interplay betwee
n cross-feeding and competition shapes microbial ecosystems and (iii) the
emergence of chaos in the ecosystems with non-reciprocal interaction.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/34/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Andrew Stuart (California Institute of Technology)
DTSTART;VALUE=DATE-TIME:20240521T161500Z
DTEND;VALUE=DATE-TIME:20240521T171500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/35
DESCRIPTION:Title: Operator Learning: Algorithms\, Analysis and Application
s\nby Andrew Stuart (California Institute of Technology) as part of No
rthwestern Applied Mathematics Seminar\n\nLecture held in M416 Tech Northw
estern Evanston IL.\n\nAbstract\nApproximating operators that map between
function spaces can be useful for accelerating systems level tasks in scie
ntific computing\, and for discovering computational models from data. In
its most basic form\, learning an operator may be cast as a form of superv
ised learning in which the input-output pairs are functions. The talk will
overview a variety of specific approximation architectures that have been
developed in the last five years\; emerging theoretical results explainin
g the approximation capabilities of the architectures will be explained\;
and applications to constitutive modeling (plasticity) and inverse problem
s (fluid mechanics) will be given.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/35/
END:VEVENT
BEGIN:VEVENT
SUMMARY:James Fitzgerald (Northwestern University)
DTSTART;VALUE=DATE-TIME:20240305T171500Z
DTEND;VALUE=DATE-TIME:20240305T181500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/36
DESCRIPTION:Title: Bridging Theoretical Neuroscience and Neural Data Scienc
e with Simple Models\nby James Fitzgerald (Northwestern University) as
part of Northwestern Applied Mathematics Seminar\n\nLecture held in M416
Tech Northwestern Evanston IL.\n\nAbstract\nNeuroscience is entering a dat
a-rich era that requires computational approaches and opens exciting new p
ossibilities for theory building. Here I will describe how my research gro
up couples neural data science and abstract theoretical modeling to build
interpretable models of diverse neural systems. In my first illustrative e
xample\, I’ll explain our efforts to uncover synaptic plasticity mechani
sms that allow flies to learn probabilistic reward associations and direct
decision making. I’ll then shift to larval zebrafish and whole-brain im
aging to explain how we build and analyze neural network models of sensori
motor processing and behavior.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/36/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Peko Hosoi (Massachusetts Institute of Technology (MIT))
DTSTART;VALUE=DATE-TIME:20240327T210000Z
DTEND;VALUE=DATE-TIME:20240327T220000Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/37
DESCRIPTION:Title: A Few Short Stories about Probability and Sports - Reiss
Lecture\nby Peko Hosoi (Massachusetts Institute of Technology (MIT))
as part of Northwestern Applied Mathematics Seminar\n\nLecture held in M41
6 Tech Northwestern Evanston IL.\n\nAbstract\nIn most professional sports\
, every physical attribute of an athlete that can be measured is tracked a
nd recorded. There exists an abundance of (relatively) high quality data
— in football\, basketball\, baseball\, cricket\, etc. — which makes s
ports an ideal testing ground for new analyses and algorithms. In this tal
k I will describe a few studies that lie at the intersection of sports and
data. Topics may include: the origin of the increase in home runs in Majo
r League Baseball\; the public health impact of allowing fans in American
football stadiums during the pandemic\; the role of skill and chance in sp
orts and other activities\; measuring “court sense” i.e. an athlete’
s decision-making ability in basketball\; and the design of optimal runnin
g shoes.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/37/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Maciej Lisicki (University of Warsaw and UPenn)
DTSTART;VALUE=DATE-TIME:20240409T161500Z
DTEND;VALUE=DATE-TIME:20240409T171500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/38
DESCRIPTION:Title: Tales of tails: Elastohydrodynamics of microscale motion
\nby Maciej Lisicki (University of Warsaw and UPenn) as part of Northw
estern Applied Mathematics Seminar\n\nLecture held in M416 Tech Northweste
rn Evanston IL.\n\nAbstract\nA look into the microworld reveals plethora o
f swimming microorganisms\, which display a rich variety of shapes and swi
mming gaits. Despite this diversity\, physics of microscale imposes univer
sal limitations on their propulsion strategies. In my talk\, I will review
the basic properties of Stokes flows and their consequences on swimming.
Next\, I will show an artificial system of microscale oil droplets that ha
ve the ability to swim due to a surface phase transition driven by environ
mental temperature fluctuations. I will demonstrate how a coarse-grained e
lastohydrodynamic model can be successfully employed to quantitatively des
cribe the motion of droplets. I will also show a couple of other examples
where a simplified elastohydrodynamic model proves useful for the predicti
on of diffusive properties.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/38/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Naomi Oppenheimer (Tel Aviv University\, Israel)
DTSTART;VALUE=DATE-TIME:20240312T180000Z
DTEND;VALUE=DATE-TIME:20240312T190000Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/39
DESCRIPTION:Title: Hydrodynamic Hamiltonians of Active Two-dimensional Flui
ds\nby Naomi Oppenheimer (Tel Aviv University\, Israel) as part of Nor
thwestern Applied Mathematics Seminar\n\nLecture held in M416 Tech Northwe
stern Evanston IL.\n\nAbstract\nI will describe two biologically inspired
systems that can be analyzed using the same hydrodynamic Hamiltonian forma
lism. The first is ATP synthase proteins\, which rotate in a biological me
mbrane. The second is swimming micro-organisms such as bacteria or algae c
onfined to a two-dimensional film. I will show that in both cases\, the ac
tive systems self-assemble into distinct structural states --- the rotatin
g proteins rearrange into a hexagonal lattice\, whereas the micro-swimmers
evolve into a zig-zag configuration with a particular tilt. While the two
systems differ both on the microscopic\, local interaction\, as well as t
he emerging\, global structure\, their dynamics originate from similar geo
metrical conservation laws applicable to a broad class of fluid flows. I w
ill then show experiments and simulations in which the Hamiltonian is pert
urbed\, leading to different and surprising steady-state configurations.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/39/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Tom Dean (Breakthrough Energy)
DTSTART;VALUE=DATE-TIME:20240423T161500Z
DTEND;VALUE=DATE-TIME:20240423T171500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/40
DESCRIPTION:Title: Contrail Climate Impacts: Modeling and Mitigation Strate
gies\nby Tom Dean (Breakthrough Energy) as part of Northwestern Applie
d Mathematics Seminar\n\nLecture held in M416 Tech Northwestern Evanston I
L.\n\nAbstract\nAviation accounts for approximately 3.5% of global anthrop
ogenic climate forcing. Of this\, less than half is attributed to the CO2
output from fuel combustion. Nearly half of the total climate impact of av
iation can be attributed to contrail cirrus – cirrus clouds that have be
en seeded by the exhaust plumes of the aircraft – though with considerab
le uncertainty. Given that only around 2-3% of all flights were are likely
responsible for 80% of the global annual contrail climate forcing\, a pot
ential contrail mitigation strategy could involve re-routing this subset o
f flights to minimize the formation of strongly warming contrails. Studies
have shown that such strategies would require as little as 0.2% additiona
l fuel burn. Implementing such a strategy requires an accurate ability to
forecast contrail impacts\, which in turn requires accurate modeling of m
icrophysical atmospheric processes at flight levels. In this talk we will
give an overview of contrail modeling and describe the grid-based Contrail
Cirrus Prediction model (CoCiP)\, whose output is designed to be input as
a layer in flight planning tools.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/40/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Rony Granek (Ben Gurion University of the Negev)
DTSTART;VALUE=DATE-TIME:20240514T161500Z
DTEND;VALUE=DATE-TIME:20240514T171500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/41
DESCRIPTION:Title: Active Fractal Networks with Stochastic Force Monopoles
and Force Dipoles Unravel Subdiffusion of Chromosomal Loci\nby Rony Gr
anek (Ben Gurion University of the Negev) as part of Northwestern Applied
Mathematics Seminar\n\nLecture held in M416 Tech Northwestern Evanston IL.
\n\nAbstract\nMotivated by the well-known fractal packing of chromatin\, w
e study the Rouse-type dynamics of elastic fractal networks with embedded\
, stochastically driven\, active force monopoles and force dipoles that ar
e temporally correlated. We compute\, analytically – using a general the
oretical framework – and via Langevin dynamics simulations\, the mean sq
uare displacement (MSD) of a network bead. Following a short time superdif
fusive behavior\, force monopoles yield anomalous subdiffusion with an exp
onent identical to that of the thermal system. In contrast\, force dipoles
do not induce subdiffusion\, and the early superdiffusive MSD crosses ove
r to a relatively small\, system-size-independent saturation value. In add
ition\, we find that force dipoles may lead to ``crawling" rotational moti
on of the whole network\, reminiscent of that found for triangular micro-s
wimmers and consistent with general theories of the rotation of deformable
bodies. Moreover\, force dipoles lead to network collapse beyond a critic
al force strength\, which persists with increasing system size\, signifyin
g a true first-order dynamical phase transition. We apply our results to t
he motion of chromosomal loci in bacteria and yeast cells' chromatin\, whe
re anomalous sub-diffusion\, MSD∼ t^ν with ν≃ 0.4\, were found in bo
th normal and ATP-depleted cells\, albeit with different apparent diffusio
n coefficients. We show that the combination of thermal\, monopolar\, and
dipolar forces in chromatin is typically dominated by the active monopolar
and thermal forces\, explaining the observed normal cells vs the ATP-depl
eted cells behavior.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/41/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Krishna Shrinivas (Northwestern University)
DTSTART;VALUE=DATE-TIME:20241008T161500Z
DTEND;VALUE=DATE-TIME:20241008T171500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/42
DESCRIPTION:Title: The Many Phases of a Cell\nby Krishna Shrinivas (Nor
thwestern University) as part of Northwestern Applied Mathematics Seminar\
n\nLecture held in M416 Tech Northwestern Evanston IL.\n\nAbstract\nCells
routinely orchestrate reactions\, interactions\, and transport amongst bil
lions of biomolecules in a crowded environment to perform the diverse task
s that underpin life. Rather than occurring in a well-mixed milieu\, biomo
lecules self-organize into dozens of membrane-lacking compartments called
condensates that enable key biological functions and are aberrant in disea
se. I will introduce how phase transitions are emerging as a paradigm unde
rlying condensate assembly and function in cells. During the talk\, I will
describe our efforts that bridge statistical physics\, applied mathematic
s\, and computation to predict emergent multiphase behavior in highly mult
icomponent soft materials and how such materials can be designed to perfor
m information processing or computational tasks.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/42/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Trevor GrandPre (Princeton University)
DTSTART;VALUE=DATE-TIME:20241015T161500Z
DTEND;VALUE=DATE-TIME:20241015T171500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/43
DESCRIPTION:Title: Model-free Inference from Time-series Data\nby Trevo
r GrandPre (Princeton University) as part of Northwestern Applied Mathemat
ics Seminar\n\nLecture held in M416 Tech Northwestern Evanston IL.\n\nAbst
ract\nDrawing inferences from experimental data often involves imposing mo
dels\, which can lead to inaccurate conclusions. Theory sometimes points t
o quantities that are significant independent of the underlying mechanisms
\, but making accurate model-free estimates of these quantities can be har
d because finite data generates systematic errors. I present two cases whe
re we develop new methods to address and correct these errors: (1) extract
ing long-term population growth from single-cell lineage data and (2) esti
mating the evidence for the arrow of time in patterns of neural activity.
For population growth\, key observables are the number of divisions and ge
neration times along a lineage for a fixed time. Simple growth rate estima
tors suffer from finite-time bias at short times\; this bias scales invers
ely with time and can be corrected. At longer times\, rare events introduc
e a linearization bias\, causing an abrupt phase transition explained by a
simple model of disordered systems. Our approach yields accurate estimate
s provided the lineage counts and lengths stay below the critical point\,
allowing inference of how mutations and physiological variations impact fi
tness. In the case of neural activity\, the relevant observables are the m
oments of activity and the waiting times between these moments. Estimating
the irreversibility—quantified by the Kullback-Leibler divergence betwe
en the distribution of forward and backward trajectories—faces similar b
iases. Finding the systematic dependence of these biases on sample size al
lows for accurate estimates\, including detecting systems that obey detail
ed balance\, and opens a path to exploring how the brain represents the ar
row of time. Generally\, this new understanding of how model-free estimato
rs rely on a complex order of limits of the amount of data and the length
of each sample may allow quantitative understanding of other relevant proc
esses such as gene regulation\, cell-cycle dynamics\, and signal transduct
ion.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/43/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Pedram Hassanzadeh (The University of Chicago)
DTSTART;VALUE=DATE-TIME:20241029T161500Z
DTEND;VALUE=DATE-TIME:20241029T171500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/44
DESCRIPTION:Title: Integrating the Spectral Analyses of Neural Networks and
Nonlinear Physics for Explainability\, Generalizability\, and Stability\nby Pedram Hassanzadeh (The University of Chicago) as part of Northwest
ern Applied Mathematics Seminar\n\nLecture held in M416 Tech Northwestern
Evanston IL.\n\nAbstract\nIn recent years\, there has been substantial int
erest in using deep neural networks (NNs) to improve the modeling and pred
iction of complex\, multiscale\, nonlinear dynamical systems such as turbu
lent flows and Earth’s climate. In idealized settings\, there has been s
ome progress for a wide range of applications from data-driven spatio-temp
oral forecasting to long-term emulation to subgrid-scale modeling. However
\, to make these approaches practical and operational\, i.e.\, scalable to
real-world problems\, a number of major questions and challenges need to
be addressed. These include 1) instabilities and the emergence of unphysic
al behavior\, e.g.\, due to how errors amplify through NNs\, 2) learning i
n the small-data regime\, 3) interpretability based on physics\, and 4) ou
t-of-distribution generalization (e.g.\, extrapolation to different parame
ters\, forcings\, and regimes) which is essential for applications to non-
stationary systems such as a changing climate. While some progress has bee
n made in addressing (1)-(4)\, e.g.\, doing transfer learning for generali
zation\, these approaches have been often ad-hoc\, as currently there is n
o rigorous framework to analyze deep NNs and develop systematic and genera
l solutions to (1)-(4). In this talk\, I will discuss some of the approach
es to address (1)-(4)\, for example\, once we identify spectral bias as th
e cause of instabilities in state-of-the-art weather models like Pangu-wea
ther\, GraphCast\, and FourCastNet. Then I will introduce a new framework
that combines the spectral (Fourier) analyses of NNs and nonlinear physics
\, and leverages recent advances in theory and applications of deep learni
ng\, to move toward rigorous analysis of deep NNs for applications involvi
ng dynamical systems. For example\, this approach can guide and explain tr
ansfer learning and pruning in such applications. I will use examples from
turbulence modeling and weather/climate prediction to discuss these metho
ds and ideas.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/44/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Miranda Holmes-Cerfon (The University of British Columbia)
DTSTART;VALUE=DATE-TIME:20241105T171500Z
DTEND;VALUE=DATE-TIME:20241105T181500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/45
DESCRIPTION:Title: The Dynamics of Particles with Ligand-receptor Contacts<
/a>\nby Miranda Holmes-Cerfon (The University of British Columbia) as part
of Northwestern Applied Mathematics Seminar\n\nLecture held in M416 Tech
Northwestern Evanston IL.\n\nAbstract\nOne way to glue objects together at
the nanoscale or microscale is by ligand-receptor interactions\, where sh
ort sticky hair-like ligands stick to receptors on another surface\, much
like velcro on the nanoscale. Such interactions are common in biological s
ystems\, such as white blood cells\, virus particles\, cargo in the nuclea
r pore complex\, etc\, and they are also useful in materials science\, whe
re coating colloids with single-stranded DNA creates particles with progra
mmable interactions. In these systems\, the ligand-receptor interactions n
ot only hold particles together\, but also influence their dynamics. How d
o such particles move? Do they “roll” on each others’ surfaces\, as
is commonly thought? Or could they slide? And does it matter? In this talk
I will introduce our modelling and experimental efforts aimed at understa
nding the coarse-grained dynamics of particles with ligand-receptor intera
ctions. Our models predict these interactions can change the particles' ef
fective diffusion by orders of magnitude. Our experiments\, using DNA-coat
ed colloids\, verify this dramatic dynamical slowdown\, but also show othe
r dynamical features not yet captured by our models\, which suggest new av
enues for exploration.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/45/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Mariela Petkova (Harvard University)
DTSTART;VALUE=DATE-TIME:20241112T171500Z
DTEND;VALUE=DATE-TIME:20241112T181500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/46
DESCRIPTION:Title: Carving Circuits from Correlative Connectomics\nby M
ariela Petkova (Harvard University) as part of Northwestern Applied Mathem
atics Seminar\n\nLecture held in M416 Tech Northwestern Evanston IL.\n\nAb
stract\nTracing the wiring diagram of a neural circuit is a powerful appro
ach to test\ncomputational models of brain function. Connectomics – the
mapping of\nneurons and their connections in the brain – poses formidabl
e technical and\nconceptual challenges. The former entails the acquisition
and reduction of\nvast amounts of electron microscopy image data to conne
ctivity matrices\,\nachieved through multi-year\, multi-team collaborative
efforts. The latter\nchallenge is the problem of coarse-graining the conn
ectivity matrices to\nneural circuits which exhibit neural activity and fu
nction. I will describe both\nchallenges in the context of mapping the ner
vous system of a small\nvertebrate – the larval zebrafish. In the same a
nimal\, we pair electron\nmicroscopy with functional information from ligh
t microscopy to interrogate\nneural circuit models for animal behavior fro
m sensory inputs to motor output.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/46/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Ertugrul Ozbudak (Northwestern University)
DTSTART;VALUE=DATE-TIME:20241119T171500Z
DTEND;VALUE=DATE-TIME:20241119T181500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/47
DESCRIPTION:Title: Reengineering Somite Segmentation without the Vertebrate
Segmentation Clock\nby Ertugrul Ozbudak (Northwestern University) as
part of Northwestern Applied Mathematics Seminar\n\nLecture held in M416 T
ech Northwestern Evanston IL.\n\nAbstract\nSomitogenesis establishes the s
egmental pattern of the vertebral column. A molecular segmentation clock s
ets the pace of somite formation while a spatial gradient of ERK activity
instructs segment boundary positions. However\, how cells are primed to fo
rm a segment boundary at discrete locations and how the clock and gradient
are mechanistically integrated were unclear. We developed transgenic repo
rters for the segmentation clock and the gradient in zebrafish embryos. We
show that the segmentation clock periodically inhibits the gradient\, the
refore projecting its oscillation onto the gradient. Pulsatile inhibition
of the gradient can fully substitute for the role of the clock\, and other
targets of the clock are dispensable for sequential segmentation. We prop
ose a “Clock-dependent Oscillatory Gradient (COG)” model in which the
clock periodically triggers discrete jumps of the positional information.
Computational simulations of the COG model explain all experimental observ
ations\; the new model effectively replaces a long-standing textbook model
.\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/47/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Evelyn Tang (Rice University)
DTSTART;VALUE=DATE-TIME:20241203T171500Z
DTEND;VALUE=DATE-TIME:20241203T181500Z
DTSTAMP;VALUE=DATE-TIME:20241112T142221Z
UID:Modeling_and_Computation/48
DESCRIPTION:by Evelyn Tang (Rice University) as part of Northwestern Appli
ed Mathematics Seminar\n\nLecture held in M416 Tech Northwestern Evanston
IL.\nAbstract: TBA\n
LOCATION:https://researchseminars.org/talk/Modeling_and_Computation/48/
END:VEVENT
END:VCALENDAR