0.25
0.5
0.75
1.25
1.5
1.75
2
Sparse modeling: some unifying theory and “topic-imaging”
Published on May 06, 20114810 Views
Information technology has enabled collection of massive amounts of data in science, engineering, social science, finance and beyond. Extracting useful information from massive and high-dimensional d
Related categories
Chapter list
Some Unifying Theory and Topic Imaging00:00
David Blackwell (1919-2010)00:09
IT revolution --> data revolution01:09
Spectrum of Media Reporting03:31
Improves News Media Analysis, Improves News Media, Improves How the World Works04:10
Our Approach to Media Analysis: Topic-Imaging05:29
Case Study: Topic Image of Business Section of NYT06:16
Case Study (cont)07:15
Solving a modern data problem09:11
Today’s Talk10:06
Occam’s Razor10:48
Occam’s Razor via Model Selection in Linear Regression11:39
Sparse Modeling in the 70’s: Model Selection12:34
Model Selection for Topic-Imaging Problem15:00
Lasso: L1-norm as a Penalty17:11
Lasso: Computation and Evaluation18:15
Lasso: Theoretical Work19:54
Regularized M-estimation including Lasso23:05
Example 1: Lasso (sparse linear model)24:21
Example 2: Structured (inverse) Cov. Estimation25:49
Example 3: Low-rank matrix approximation26:11
Unified Analysis27:37
Why can we estimate parameters?29:25
In high-dim and when r corresponds to true structure (e.g. sparsity), why estimation is still possible30:33
Main Result for Regularized M-estimation33:44
Examples of decomposable regularizers34:38
Recovering Existing Result in Bickel et al 0834:58
Obtaining New Result (Robustness of Lasso)35:55
Summary of unified analysis36:50
Partial Summary37:21
Topic Imaging: Subject-Specific Summarization of Document Corpus37:59
Our approach: predictive sparse methods + human experiment38:26
Sample Result from Our Approach (Document-S^3)39:10
Document Corpus40:19
Pre-processing40:56
Matrix Set-up and Labeling41:18
List of Generation Methods41:38
Flow Chart of Our Automatic Summarization42:00
How to Select from 120 Possible Lists for One Topic?42:10
Prediction Is Not the Goal43:17
A Human Experiment44:26
Running Human Experiment45:18
Human Experiment Results in a Glance46:35
Qualitative Summary of Human Exp. Results47:25
Summary of talk48:12
Future Directions: Richer Data/Subejct Applications49:36
Future Directions: Research Topics50:00
Acknowledgements50:54
Stat-news project (El Ghaoui and Yu)51:21