Statistical Inference for Large Matrices under Complexity Constraints
The purpose of this working group is to discuss important problems related to statistical inference for large matrices under various complexity constraints and to try to make progress on some of them. The problems in question include matrix completion, more general problems of matrix estimation based on linear measurements, covariance matrix estimation, estimation of density matrix in quantum state tomography, matrix recovery in compressed sensing, hypotheses testing for large matrices, etc. The complexity constraints include low rank properties, sparsity of the matrix in a given dictionary, smoothness assumptions for kernels on weighted graphs or manifolds, and various combinations of these complexity assumptions. The analysis of these problems relies on a variety of mathematical tools in high-dimensional probability and asymptotic geometric analysis developed in the recent years. In particular, the methods of nonasymptotic theory of random matrices are increasingly important in matrix estimation. The models based on log-concave distributions in high-dimensional spaces could be also of importance (in addition to more classical Gaussian and subgaussian models).
Working group meeting #14
Greetings,
Olga Klopp from the Paris West University Nanterre La Défense will be the speaker this Wednesday during our working group meeting.
She will present some new results on 1-bit matrix completion and possible direction for future research.
Please join us for the discussion!
To join the online meeting (Now from mobile devices!)
-------------------------------------------------------
1. Go to https://samsi.webex.com/samsi/j.php?MTID=m37b745bdec83580124c48af7db6d524d
2. If requested, enter your name and email address.
Working group meeting #13
Greetings,
Next Wednesday will be our first group meeting of 2014. We are pleased to have Peter Kim as our speaker. The title and abstract of his talk are below.
Title: Testing quantum states for purity
Working group meeting #12
Greetings,
Topic: Statistical Inference for Large Matrices under Complexity Constraints
Date: Wednesday, November 27, 2013
Time: 3:00 pm, Eastern Standard Time (New York, GMT-05:00)
Meeting Number: 688 706 951
Meeting Password: Nuclear1
Our next group meeting is this Wednesday (November, 27). The meeting time can be used for private communication between members.
-------------------------------------------------------
To join the online meeting
-------------------------------------------------------
Working group meeting #11
Hello,
_______________________________________________________________________
Please note that the meeting will start at 2:30 pm instead of 3 pm.
_______________________________________________________________________
This week, we are pleased to welcome our guest speaker, Shahar Mendelson from Technion and Australian National University. The title of Shahar's talk is
"Random matrices and the way they act on subsets of R^n".
Abstract:
I will present results on the structure of random images of subsets of R^n,
Working group meeting #10
Greetings,
Our next group meeting is this Wednesday (November, 13). We don't have a speaker, so the meeting time can be used for private communications between members.
Topic: Statistical Inference for Large Matrices under Complexity Constraints
Date: Wednesday, November 13, 2013
Time: 3:00 pm, Eastern Standard Time (New York, GMT-05:00)
Meeting Number: 688 179 013
Meeting Password: Nuclear1
-------------------------------------------------------
To join the online meeting (Now from mobile devices!)
-------------------------------------------------------
Working group meeting #9
Hello,
This week, Ming Yuan will talk about "Optimal Detection of Sparse Signal Segments". The abstract is below. Please note that the meeting will start at 2:30 pm instead of 3pm.
____________________________________________________________________________
Ming Yuan, Department of Statistics, University of Wisconsin-Madison.
Title: Optimal Detection of Sparse Signal Segments
Abstract:
Working group meetings 8 and 9
Hello,
This week, our working group meeting time will be used for private discussions between members.
The following Wednesday (November, 6), Ming Yuan will talk about "Optimal Detection of Sparse Signal Segments". Please note that his talk will start at 2:30pm instead of 3pm.
I will post a more detailed announcement before next meeting.
-------------------------------------------------------
Topic: Statistical Inference for Large Matrices under Complexity Constraints
Date: Wednesday, October 30, 2013
Time: 3:00 pm, Eastern Daylight Time (New York, GMT-04:00)
Working group meeting #7
Greetings,
Our next group meeting is tomorrow (October, 23). We don't have a speaker, so the meeting time can be used for private communications between members.
If you want to present your work, discuss an interesting paper or suggest a problem to potential collaborators one of the following weeks, please let me know.
If you are already working on a project with a smaller group of people, the meeting time can be reserved for your private communication.
To join the online meeting (Now from mobile devices!)
-------------------------------------------------------
Working group meeting #6
Greetings,
This Wednesday, we are pleased to have Vladimir Koltchinskii as a speaker. He will talk about estimation of low rank smooth kernels on weighted graphs. The title and abstract are below.
Title:
Estimation of Low Rank Smooth Kernels on Weighted Graphs.
Abstract:
We discuss a version of noisy matrix completion problem in
which the goal is to estimate a symmetric kernel defined on the vertex set of
a weighted graph based on i.i.d. noisy observations of the kernel at randomly
picked couples of vertices of the graph. The underlying assumption is that
Working group meeting #5
Greetings,
Please join our next discussion tomorrow at 3 pm Eastern time.
Topic: Statistical Inference for Large Matrices under Complexity Constraints
Date: Wednesday, October 9, 2013
Time: 3:00 pm, Eastern Daylight Time (New York, GMT-04:00)
Meeting Number: 680 457 029
Meeting Password: Nuclear1
-------------------------------------------------------
To join the online meeting (Now from mobile devices!)
-------------------------------------------------------