Semi-Random Models and Robustness in Unsupervised Learning
Talkat Robust and High-Dimensional Statistics (Simons Institute workshop, 2018) · 2018-10-29..11-02·
summary
A semi-random model mixes a tiny fraction beta of truly random sparse supports with adversarially chosen supports and shows polynomial-time recovery of the dictionary A and sparse codes X up to any accuracy when incoherence holds and sparsity up to sqrt(n) (purely random analysis extends to about n^{2/3}). Recovery uses a testing procedure that reads inner-product histograms (bimodal if a candidate equals a column) plus weighted-sum candidate generation, proven via coupling/CLT and concentration bounds; identifiability follows from a mild triples condition.
AI-written summary · about
chapters (16)
- 0:00Semi-random Models and Robustness in Unsupervised Learning
- 0:31Probabilistic Modeling in ML
- 1:33Semirandom models
- 4:19Monotone Adversaries Feige-Kilian
- 5:22Monotone Adversaries - 1" type **Improving" planted solution
- 8:06Monotone Adversaries - 2nd type Additional information about signal
- 15:57Guarantees for Dictionary Learning
- 20:56Goal
- 26:58Main Results
- 29:40o(n2/3) Sparsity in the Random Model (B = 1)
- 32:09Identifiability Under Minimal Assumptions
- 33:23Techniques
- 34:52Testing Procedure
- 38:30Key Lemma (Soundness for NO case)
- 40:59Candidate Generation
- 44:25Conclusion
related talks
45mHigh Dimensional Robust Sparse Regression
12mDistribution-Free Robustness
01h 05mRecovering Hidden Sparsity via Sum of Squares
40mThe Trimmed Lasso: Sparse Recovery Guarantees And Practical Optimization
01h 01mRobustness, Stochastics, Uncertainty 1
53mDesigning Robust Learners
19mSupport Recovery in Universal One-bit Compressed Sensing
01h 24mAdversarial Examples
