Difference between revisions of "Algebra in Statistics and Computation Seminar"

From UW-Math Wiki
Jump to: navigation, search
(3 intermediate revisions by the same user not shown)
Line 1: Line 1:
'''When''': 1:30 pm, Thursdays
+
'''When''': 1:30-2:25, Thursdays (1:30-1:45pm Social Chit-Chat, 1:45-2:25 Talk Time)
  
'''Where''': Virtual
+
'''Where''': Virtual: https://uwmadison.zoom.us/j/95934501565
  
 
'''Contact''': [https://www.math.wisc.edu/~jose/  Jose Israel Rodriguez], [https://sites.google.com/wisc.edu/zinanwang/ Zinan Wang] (Lead)
 
'''Contact''': [https://www.math.wisc.edu/~jose/  Jose Israel Rodriguez], [https://sites.google.com/wisc.edu/zinanwang/ Zinan Wang] (Lead)
Line 10: Line 10:
  
 
{| cellpadding="8"
 
{| cellpadding="8"
!align="left" | date
+
!align="left" | Date
!align="left" | speaker
+
!align="left" | Speaker
!align="left" | title
+
!align="left" | Title
!align="left" | host(s)
+
!align="left" | Host
 
|-  
 
|-  
 
|February 11
 
|February 11
 
|[https://www.math.ubc.ca/~erobeva/ Elina Robeva (UBC)]
 
|[https://www.math.ubc.ca/~erobeva/ Elina Robeva (UBC)]
|TBD
+
|Hidden Variables in Linear Causal Models
|Local
 
 
|-
 
|-
 
|March 11
 
|March 11
 
|[http://www.luke-amendola.appspot.com/ Carlos Amendola (ULM)]
 
|[http://www.luke-amendola.appspot.com/ Carlos Amendola (ULM)]
 
|TBD
 
|TBD
|Local
+
|
 
|-
 
|-
 
|April 8
 
|April 8
 
|[https://people.maths.ox.ac.uk/seigal/ Anna Seigal (Oxford)]
 
|[https://people.maths.ox.ac.uk/seigal/ Anna Seigal (Oxford)]
 
|TBD
 
|TBD
|Local
+
|
 
|-
 
|-
 
|}
 
|}
  
 
== Abstracts ==
 
== Abstracts ==
 +
===February 11: Elina Robeva===
 +
Title: Hidden Variables in Linear Causal Models.
  
 +
Abstract: Identifying causal relationships between random variables from observational data is an important hard problem in many areas of data science. The presence of hidden variables, though quite realistic, pauses a variety of further problems. Linear structural equation models, which express each variable as a linear combination of all of its parent variables, have long been used for learning causal structure from observational data. Surprisingly, when the variables in a linear structural equation model are non-Gaussian the full causal structure can be learned without interventions, while in the Gaussian case one can only learn the underlying graph up to a Markov equivalence class. In this talk, we first discuss how one can use high-order cumulant information to learn the structure of a linear non-Gaussian structural equation model with hidden variables. While prior work posits that each hidden variable is the common cause of two observed variables, we allow each hidden variable to be the common cause of multiple observed variables. Next, we discuss hidden variable Gaussian causal models and the difficulties that arise with learning those. We show it is hard to even describe the Markov equivalence classes in this case, and we give a semi algebraic description of a large class of these models.
 
----
 
----
 +
== Other events to note ==
 +
{| cellpadding="8"
 +
!align="left" | date
 +
!align="left" | event/title
 +
!align="left" | speaker
 +
|-
 +
|Fourth Thursday's of the month
 +
|[https://www.math.wisc.edu/wiki/index.php/Applied_Algebra_Seminar Applied Algebra Seminar]
 +
|
 +
|-
 +
|}

Revision as of 23:54, 31 January 2021

When: 1:30-2:25, Thursdays (1:30-1:45pm Social Chit-Chat, 1:45-2:25 Talk Time)

Where: Virtual: https://uwmadison.zoom.us/j/95934501565

Contact: Jose Israel Rodriguez, Zinan Wang (Lead)

Remark: This informal seminar is held on the second Thursday of the month

Spring 2021 Schedule

Date Speaker Title Host
February 11 Elina Robeva (UBC) Hidden Variables in Linear Causal Models
March 11 Carlos Amendola (ULM) TBD
April 8 Anna Seigal (Oxford) TBD

Abstracts

February 11: Elina Robeva

Title: Hidden Variables in Linear Causal Models.

Abstract: Identifying causal relationships between random variables from observational data is an important hard problem in many areas of data science. The presence of hidden variables, though quite realistic, pauses a variety of further problems. Linear structural equation models, which express each variable as a linear combination of all of its parent variables, have long been used for learning causal structure from observational data. Surprisingly, when the variables in a linear structural equation model are non-Gaussian the full causal structure can be learned without interventions, while in the Gaussian case one can only learn the underlying graph up to a Markov equivalence class. In this talk, we first discuss how one can use high-order cumulant information to learn the structure of a linear non-Gaussian structural equation model with hidden variables. While prior work posits that each hidden variable is the common cause of two observed variables, we allow each hidden variable to be the common cause of multiple observed variables. Next, we discuss hidden variable Gaussian causal models and the difficulties that arise with learning those. We show it is hard to even describe the Markov equivalence classes in this case, and we give a semi algebraic description of a large class of these models.


Other events to note

date event/title speaker
Fourth Thursday's of the month Applied Algebra Seminar