# Resources

I'll put my code, presentation slides, useful links, and various other useful resources here.

l

### Selected recent talk slides

- Talk on Watermarking LLMs at UCLA [slides]
- Talk on Permute-and-Flip Decoding and Watermarking for LLMs [slides]
- Open problem for DP-ERM at TTIC [slides]
- Talk on watermarking generative AI at ICML/KDD [slides]
- Talk on deep learning theory at UCSB Statistics/GeorgiaTech/USC/MBZUAI [slides]
- Talk on Offline and Low-Adaptive RL at INFORMS[slides]
- Talk on Universal Dynamic Regret at Princeton [slides]
- Talk on privacy accounting at Rutgers/LinkedIn/MIT [slides]
- Talk on the Local Adaptivity of Deep Neural Networks at Rice [slides]
- Talk on Harnessing Nonstationarity at Genetech[slides]
- Talk on dynamic pricing at UCSB [slides]
- Talk on online forecasting at JSM-2021 [slides]
- RL Theory Talk on Uniform OPE[slides]
- UCSD Talk on Online Forecasting[slides]
- Berkeley Simons Institute Talk on Subsampled RDP [slides]
- Policy learning talk at Caltech MCS / UCSB Statistics [slides]
- Talk on Per-Instance DP at UC Santa Cruz's PiG workshop [slides]
- My PhD thesis defense slides [slides]
- CMU and UCSB talks "SignSGD / Signum optimizers and deep learning with Gluon" [slides]
- NIPS'16 "What-If“ workshop Talk on "Optimal and Adaptive Off-policy Evaluation" [slides]
- ICML'16 TPDP workshop Talk on "Generalization and Learnability under Differential Privacy and its variants" [slides]
- Columbia talk on GTF and "Total variation class beyond 1D" [slides]
- ML Lunch talk on "Falling factorials" and "Graph Trend Filtering" [slides]
- "Privacy for free talk at ICML'15" [slides]

### Code and software

- Trend filtering [github]
- Trend filtering on Graphs[demo code, uci experiments (somewhat messy), wavelets baselines]
- Falling factorial basis transform [demo code]
- Ecological inference via distribution regression (Thanks, Dougal!)[github]
- Fast differentially private matrix factorization [github]
- Codes to run OPS, SGLD, SGNHT to differentially private learning from our ICML'15 paper. [demo code]
- PARSuMi for robust matrix completion with sparse corruptions. [demo code]
- Lasso-SSC and LRSSC for subspace clustering from our ICML'13 / NIPS'13 papers. [demo code]

### Some really old journal club/discussion class presentations back in Singapore

- "Stochastic Subgradient Descent for Nuclear Norm Regularization" [slides]
- Non-negative Matrix Factorization(NMF) and "A-Optimal Non-Negative Projection for Image Representation" [slides]
- Differential privacy tutorial [slides]
- Deep Learning and "Sparse modeling of human actions from motion imagery" [slides]
- Subspace Clustering with missing data: "High rank matrix completion" [slides]
- Discussion of Wiberg L1 (CVPR10 Best Paper) [slides]

### Learning deep learning with Mxnet Gluon

A book made of jupyter notebooks on github. All examples are runnable. mxnet Gluon is highly flexible and very suitable for researchers. 中文版

### Adventures in Data Land

Alex's blog with many ideas and practical tricks for using ML.

### Csaba Szepesvari's blog on Bandits

Mostly stochastic and adversarial linear bandits.

### Larry Wasserman's blog: Normal Deviate

Cool blog from the CMU statistician on statistics and machine learning topics.

### John Langford's blog: hunch.net

A renowned machine learning theory blog. A few good/interesting posts per month. To have a flavor, check out the article: Adversarial Academia.

### Matrix Factorization Jungle

A comprehensive site that keeps updating the state-of-the-art algorithms, theory and evaluations in MF related fields, including: Matrix Completion, Matrix Recovery(RPCA), Compressive Sensing, Dictionary learning, Non-negative Matrix Factorization and etc.

### Nuit Blanche blog on Compressive Sensing and Matrix Factorization

The maintainer of Matrix Factorization Jungle (Igor Carron), articles are faster than updates on the summary site.

### Compressive sensing resources

An almost thorough list of compressive sensing papers, reviews and tutorials.

### Ma Yi's Low-rank matrix recovery & completion page

A list of papers on nuclear norm based convex methods for low-rank matrix. Useful code samples of Augmented Lagrange Multiplier methods for RPCA.