Robust Estimation and Generative Adversarial Nets

on YouTube ↗44m

" summary

Works through Huber epsilon contamination and shows that depth estimators like Tukey's median and multi-task regression depths achieve exact minimax rates for location, regression and covariance problems but are computationally intractable beyond p≈10. Casts depth as total variation F‑learning so GANs become optimization proxies. Pure TV‑GAN fails in practice, while JS‑GAN discriminators with hidden layers provably recover nonlinear moments and give robust estimators that scale to p≈5000. Practical caveat is unstable nonconvex optimization and simple fixes for covariance (intercept neuron or extra layer) are required.

AI-written summary · about

" chapters (12)

1. 0:00Intro
2. 0:52Huber's Model
3. 3:19An Example
4. 4:45Multivariate Location Depth
5. 8:06Multi-task Regression Depth
6. 12:55Covariance Matrix
7. 15:24Summary
8. 17:58Computational Challenges
9. 18:18Advantages of Tukey Median
10. 19:26f-Learning
11. 26:17theoretical foundation
12. 31:54JS-GAN

" related talks

• 11mNonparametric Density Estimation and Convergence of GANs under Besov IPMsAnanya Uppal · Meet the Fellows Welcome Event
• 47mIs Minimax Optimization the Right Framework to Understand Learning in Minimax Games?Gauthier Gidel · Adversarial Approaches in Machine Learning 2022
• 01h 00mTutorial on Deep Learning IVRuslan Salakhutdinov · Foundations of Machine Learning Boot Camp 2017
• 01h 30mDeep Learning Techniques for Monocular Depth EstimationVladimir Kovacevic · DSC Europe 23
• 52mDeep Generative Models And Unsupervised Methods For Inverse ProblemsAlex Dimakis · Algorithmic Advances for Statistical Inference with Combinatorial Structure
• 40mTraining generative models from locally privatized data via entropic optimal transportAyfer Ozgur · Trust in Decentralized Systems 2026
• 43mHitchhiker’s Guide to Generative Adversarial Networks (GANs)Ramanan Balakrishnan · Anthill Inside 2017
• 49mGeneralized Energy-Based ModelsArthur Gretton · Adversarial Approaches in Machine Learning 2022