Adversarial Approaches in Machine Learning 2022
2022-02-17..2022-02-24
-- online --
[Quantifying Uncertainty: Stochastic, Adversarial, and Beyond, Mathematics of Online Decision Making 2020, Frontiers of Deep Learning 2019, Optimization, Statistics and Uncertainty, Optimization and Decision-Making Under Uncertainty (Simons Institute, September 2016), Data-Driven Decision Processes Boot Camp 2022]
who it's for
min-max optimization researchers, machine learning researchers, theoretical computer scientists, reinforcement learning researchers, adversarial robustness researchers, causal and semiparametric inference researchers, representation learning researchers
topics
constrained min-max optimization; extra-gradient methods; Halpern iteration for monotone operators; variational method of moments; offline reinforcement learning with density ratios; computational complexity of min-max (PPAD); dynamical systems view of algorithms; zeroth-order convex-concave methods; imitation learning via moment matching; certified robustness and perceptual threat models; generalized energy-based models; policy gradient estimation and generalization
| title | author | length |
|---|---|---|
| Last-Iterate Convergence in Constrained Min-Max Optimization: SOS to the Rescue | Yang Cai | 01h 04m |
Proves that the extra-gradient method with constant step η = C / L achieves last-iterate convergence for monotone variational inequalities over arbitrary… | ||
| Fast Rates for No-Regret Learning in General Games | Noah Golowich | 49m |
Shows that when all players use Optimistic Hedge in finite general‑sum games each player’s external regret falls to roughly O(log |A| · log^4 T) — i.e., near… | ||
| The Variational Method of Moments | Nathan Kallus | 57m |
Proposes the variational method of moments (VMM): a minimax formulation that replaces K test functions with a rich RKHS or neural-net class and an… | ||
| Offline Reinforcement Learning With Only Realizability | Jason Lee | 48m |
Proposes solving offline RL by solving a regularized primal-dual LP (density-regularized DICE-style objective) to recover an importance weight w and extract a… | ||
| Min-max Optimization: From Complexity to Algorithms | Emmanouil Zampetakis | 51m |
Shows that finding an approximate local min‑max equilibrium is PPAD-complete and that any black‑box first‑order method needs an exponential query lower bound… | ||
| Min-Max Optimization from a Dynamical Systems Viewpoint | Panayotis Mertikopoulos | 01h 00m |
Shows that first-order min–max methods (gradient descent-ascent, extra-gradient/optimistic, proximal-point and their stochastic variants) are asymptotic… | ||
| Smooth Nonconvex Min-Max Optimization | Meisam Razaviyayn | 46m |
Gives concrete solvable cases and iteration counts rather than vague promises: when f is smooth and strongly concave in the adversary coordinate a provable… | ||
| Studying Failure Modes of Deep Learning and Ways To Fix Them | Soheil Feizi | 59m |
Shows that non-additive splitting noise (SSN) produces deterministic certificates and a 1/(2λ) L1 Lipschitz bound, doubling ImageNet-scale certified radii… | ||
| Is Minimax Optimization the Right Framework to Understand Learning in Minimax Games? | Gauthier Gidel | 47m |
Shows that Adam’s edge in GANs comes from its step-norm, not direction: reproducing Adam with a normalized SGD variant (AdaLR/AdaDIR) matches or beats Adam… | ||
| Beyond Adversarial Approaches on Distributional Robustness | Qi Lei | 42m |
Proposes using label propagation over a learned representation to fix covariate shift: train on source labels, add consistency regularization over unlabeled… | ||
| Computationally Efficient Alternatives to Nonconvex-Nonconcave Min-Max Optimization | Nisheeth Vishnoi | 57m |
Proposes a new solution concept called epsilon-greedy adversarial equilibrium that limits the maximizer to greedy paths and proves such equilibria exist for… | ||
| Halpern Iteration and Equilibria Problems | Jelena Diakonikolas | 54m |
Shows that Halpern iteration (anchored fixed-point updates plus a potential function) achieves the optimal 1/K rate for making the monotone operator norm… | ||
| Zeroth-Order Methods for Convex-Concave Minmax Problems: Learning from Strategically Generated Data | Chinmay Maheshwari | 55m |
Recasts decision dependent robustness into a finite dimensional convex concave min max via Wasserstein DRO and a strategic agent model where utility equals… | ||
| Are Single-Loop Algorithms Sufficient for Unbalanced Minimax Optimization? | Niao He | 51m |
A constructive claim: a primal‑dual lifting converts FX and HY to dual variables and yields a single‑loop first‑order algorithm (extrapolation + gradient… | ||
| Policy Gradient: Optimal Estimation, Convergence, and Generalization beyond Cumulative Rewards | Mengdi Wang | 45m |
Shows that policy gradients obey a policy-gradient Bellman equation and uses a practical double fitted iteration (regression over Q and its gradient) to get a… | ||
| Of Moments and Matching: Trade-offs and Treatments in Imitation Learning | Steven Wu | 56m |
Presents a unifying game‑theoretic recipe that casts three imitation regimes as minimax moment‑matching games (off‑Q, reward‑matching, on‑Q), proves that… | ||
| Learning Uninformative Representations | Richard Zemel | 54m |
Claims an algorithm called EIEL that infers worst-case domains from a reference model that has learned shortcuts, then alternates environment assignment… | ||
| Generalized Energy-Based Models | Arthur Gretton | 49m |
Paper argues that GAN critics contain reusable density information and turns them into a generalized energy-based model that learns an energy E and normalizer… | ||
| Rebel: Combining Deep Reinforcement Learning and Search for Imperfect-Information Games | Noam Brown | 47m |
ReBeL combines counterfactual regret minimization with deep value nets and search over public belief state rooted subgames, proving convergence to a Nash… | ||
| Online Adversarial Multicalibration And (Multi)Calibeating | Aaron Roth | 48m |
A unifying framework called online minimax multi-objective optimization converts many adversarial online goals into a zero-sum surrogate game and yields… | ||
| Adversarial Machine Learning and Instrumental Variables for Flexible Causal Modeling | Vasilis Syrgkanis | 53m |
Proposes an adversarial minimax objective that converts nonparametric instrumental-variable conditions into a loss minimization problem, proves fast, adaptive… | ||
| nothing matches | ||