Computer scientists and mathematicians use "computable", "ZFC", and "Godel incompleteness" all the time, but many schools do not offer a serious math logic class, so the usual concepts from computability theory and model theory are somewhat mysterious to math community. I personally believe that all mathematicians should know the basic knowledge from math logic (not just truth table), so I will give a short overview of math logic:
What is Turing machine? What are computable and not computable? How do logicians describe and understand the math theorems? What do axioms really want to catch (yes! ZFC as an example)? What does Godel incompleteness mean?
November 18
Algebra and Combinatorics
Tropical Point Arrangements and Valuated Matroids
Griffin Edwards - Georgia Tech
Gibson Hall 126-A3:00 PM
A matrix representing a matroid over a field records which sets of columns are dependent, and its maximal minors assign a determinant to each basis. Can a tropical matrix do both for a valuated matroid, whose bases already carry weights? I’ll show how to construct such a matrix using special valuated cocircuits as its rows. The resulting columns are tropically dependent exactly on the dependent sets of the matroid, and valuated circuits give explicit coefficients for those relations. To recover the basis weights, I’ll introduce a tropical volume motivated by the Cauchy–Binet formula.
November 19
Geometry & Topology
Persistent Homology Learning Seminar (LSC)
Rafal Komendarczyk - Tulane University
Gibson Hall 30812:30 PM
Morse Theory - continued
November 20
Applied and Computational Math
What Can Snapshots Reveal? Parameter Identifiability in a Stochastic Fountain
Lan Trinh - Tulane University
Gibson Hall 126A3:00 PM
Can static images reveal the dynamic rates and sources that generated them? Motivated by spatial transcriptomics, we investigated this question through the stochastic fountain model, in which particles enter from a localised source, diffuse, and disappear through degradation or boundary absorption. At steady state, each snapshot is a realisation of a spatial Poisson point process whose intensity solves an elliptic PDE. The observations are random particle locations or spatially aggregated counts, while the PDE solution itself remains unobserved. We introduce determining summaries that connect statistical identifiability of snapshot distributions with structural identifiability of the underlying PDE model. With diffusivity known, we establish global identifiability of production and degradation rates and source location and size from particle-location data. We then examine how spatial aggregation and domain symmetries can break identifiability and relate Fisher information to parameter sensitivities. Analytical results and numerical examples show how observation design affects both parameter identifiability and the precision of parameter estimates. These identifiability results lay the groundwork for investigating consistency and asymptotic normality of joint maximum likelihood estimators of model parameters. This is joint work with Tommaso Buvoli and Scott A. McKinley.
November 16 - November 20
November 16
Monday
no events
November 17
Tuesday
Graduate Student Colloquium
A short overview of math logic
Haoxi Hu - Tulane Math Department
Richardson Building 1083:30 PM
Computer scientists and mathematicians use "computable", "ZFC", and "Godel incompleteness" all the time, but many schools do not offer a serious math logic class, so the usual concepts from computability theory and model theory are somewhat mysterious to math community. I personally believe that all mathematicians should know the basic knowledge from math logic (not just truth table), so I will give a short overview of math logic:
What is Turing machine? What are computable and not computable? How do logicians describe and understand the math theorems? What do axioms really want to catch (yes! ZFC as an example)? What does Godel incompleteness mean?
November 18
Wednesday
Algebra and Combinatorics
Tropical Point Arrangements and Valuated Matroids
Griffin Edwards - Georgia Tech
Gibson Hall 126-A3:00 PM
A matrix representing a matroid over a field records which sets of columns are dependent, and its maximal minors assign a determinant to each basis. Can a tropical matrix do both for a valuated matroid, whose bases already carry weights? I’ll show how to construct such a matrix using special valuated cocircuits as its rows. The resulting columns are tropically dependent exactly on the dependent sets of the matroid, and valuated circuits give explicit coefficients for those relations. To recover the basis weights, I’ll introduce a tropical volume motivated by the Cauchy–Binet formula.
November 19
Thursday
Geometry & Topology
Persistent Homology Learning Seminar (LSC)
Rafal Komendarczyk - Tulane University
Gibson Hall 30812:30 PM
Morse Theory - continued
November 20
Friday
Applied and Computational Math
What Can Snapshots Reveal? Parameter Identifiability in a Stochastic Fountain
Lan Trinh - Tulane University
Gibson Hall 126A3:00 PM
Can static images reveal the dynamic rates and sources that generated them? Motivated by spatial transcriptomics, we investigated this question through the stochastic fountain model, in which particles enter from a localised source, diffuse, and disappear through degradation or boundary absorption. At steady state, each snapshot is a realisation of a spatial Poisson point process whose intensity solves an elliptic PDE. The observations are random particle locations or spatially aggregated counts, while the PDE solution itself remains unobserved. We introduce determining summaries that connect statistical identifiability of snapshot distributions with structural identifiability of the underlying PDE model. With diffusivity known, we establish global identifiability of production and degradation rates and source location and size from particle-location data. We then examine how spatial aggregation and domain symmetries can break identifiability and relate Fisher information to parameter sensitivities. Analytical results and numerical examples show how observation design affects both parameter identifiability and the precision of parameter estimates. These identifiability results lay the groundwork for investigating consistency and asymptotic normality of joint maximum likelihood estimators of model parameters. This is joint work with Tommaso Buvoli and Scott A. McKinley.