arXiv research

Paper improves worst-case regret bounds for RLSVI in reinforcement learning.

Worst-case bounds on the expected shortfall risk given only limited information on the distribution of the random variables has been studied extensively in the literature. In this paper, we develop a new worst-case bound on the expected shortfall when the univariate marginals are known exactly and additional expert inf…

In three-dimensional computational topology, the theory of normal surfaces is a tool of great theoretical and practical significance. Although this theory typically leads to exponential time algorithms, very little is known about how these algorithms perform in "typical" scenarios, or how far the best known theoretical…

New framework improves worst-case generalization bounds for stochastic optimization.

We design a general framework for answering adaptive statistical queries that focuses on providing explicit confidence intervals along with point estimates. Prior work in this area has either focused on providing tight confidence intervals for specific analyses, or providing general worst-case bounds for point estimate…

We consider the problem of learning a dictionary matrix from a number of observed signals, which are assumed to be generated via a linear model with a common underlying dictionary. In particular, we derive lower bounds on the minimum achievable worst case mean squared error (MSE), regardless of computational complexity…

Exact tail probability bounds for bounded kurtosis.

We find the exact worst-case tail probability for bounded kurtosis.

This paper studies a recent proposal to use randomized value functions to drive exploration in reinforcement learning. These randomized value functions are generated by injecting random noise into the training data, making the approach compatible with many popular methods for estimating parameterized value functions. B…

MaxMatch improves SSL with worst-case consistency for better generalization.

Study optimizes identifying the best arm with fixed rounds and Gaussian outcomes.

Bandits with Knapsacks (BwK) is a general model for multi-armed bandits under supply/budget constraints. While worst-case regret bounds for BwK are well-understood, we present three results that go beyond the worst-case perspective. First, we provide upper and lower bounds which amount to a full characterization for lo…

2D Total Variation Denoising (TVD) is a widely used technique for image denoising. It is also an important nonparametric regression method for estimating functions with heterogenous smoothness. Recent results have shown the TVD estimator to be nearly minimax rate optimal for the class of functions with bounded variatio…

We introduce a class of utility-based market makers that always accept orders at their risk-neutral prices. We derive necessary and sufficient conditions for such market makers to have bounded loss. We prove that hyperbolic absolute risk aversion utility market makers are equivalent to weighted pseudospherical scoring …

New algorithm reduces worst-case regret for heavy-tailed bandits.

We present methods for online linear optimization that take advantage of benign (as opposed to worst-case) sequences. Specifically if the sequence encountered by the learner is described well by a known "predictable process", the algorithms presented enjoy tighter bounds as compared to the typical worst case bounds. Ad…

Oracle-efficient algorithms for online learning with smoothed and hint-adversaries.

There has been a large amount of interest, both in the past and particularly recently, into the power of different families of universal approximators, e.g. ReLU networks, polynomials, rational functions. However, current research has focused almost exclusively on understanding this problem in a worst-case setting, e.g…

We prove the first nontrivial worst-case lower bounds for two closely related problems. First, $Ω(n^{3/2})$ degree-1 reductions, series-parallel reductions, and $Δ$Y transformations are required in the worst case to reduce an $n$-vertex plane graph to a single vertex or edge. The lower bound is achieved by any planar g…

Any generic closed curve in the plane can be transformed into a simple closed curve by a finite sequence of local transformations called homotopy moves. We prove that simplifying a planar closed curve with $n$ self-crossings requires $Θ(n^{3/2})$ homotopy moves in the worst case. Our algorithm improves the best previou…

Paper aims to ensure reliable detection of out-of-distribution data with certifiable worst-case guarantees.

The goal of regression and classification methods in supervised learning is to minimize the empirical risk, that is, the expectation of some loss function quantifying the prediction error under the empirical distribution. When facing scarce training data, overfitting is typically mitigated by adding regularization term…

Optimizes privacy-preserving optimization for heavy-tailed data.

Improved algorithms for stochastic linear bandits using tighter confidence sequences.

In experimental design, we are given a large collection of vectors, each with a hidden response value that we assume derives from an underlying linear model, and we wish to pick a small subset of the vectors such that querying the corresponding responses will lead to a good estimator of the model. A classical approach …

Several recent works have shown that state-of-the-art classifiers are vulnerable to worst-case (i.e., adversarial) perturbations of the datapoints. On the other hand, it has been empirically observed that these same classifiers are relatively robust to random noise. In this paper, we propose to study a \textit{semi-ran…

Hardness proof for agnostically learning halfspaces from worst-case lattice problems.

The paper tackles adversarial robustness by maximizing worst-case mutual information.

The paper tackles robust control for insurance contracts under uncertain transition rates.

Improved DP SO with large Lipschitz parameters, handling outliers and heavy-tailed data.

New expressive losses improve adversarial robustness without sacrificing accuracy.

Paper proves higher-order flow matching preserves optimality in generative modeling.

The paper analyzes extreme risk measures with limited distributional information.

In this paper, we propose the uncertain volatility models with stochastic bounds. Like the regular uncertain volatility models, we know only that the true model lies in a family of progressively measurable and bounded processes, but instead of using two deterministic bounds, the uncertain volatility fluctuates between …

Proposes a new uncertain volatility model with worst-case scenario analysis.

DRCS selects a subset of data to minimize worst-case test error under covariate shift.

We introduce a modular framework for market making. It combines cost-function based automated market makers with bandit algorithms. We obtain worst-case profits guarantee's relative to the best in hindsight within a class of natural "overround" cost functions . This combination allow us to have distribution-free guaran…

We propose an approach to the aggregation of risks which is based on estimation of simple quantities (such as covariances) associated to a vector of dependent random variables, and which avoids the use of parametric families of copulae. Our main result demonstrates that the method leads to bounds on the worst case Valu…

This paper studies bounds for the Lipschitz constant of random neural networks.

New policy optimizes risk and optimality in stochastic bandits.

Submodular extensions of an energy function can be used to efficiently compute approximate marginals via variational inference. The accuracy of the marginals depends crucially on the quality of the submodular extension. To identify the best possible extension, we show an equivalence between the submodular extensions of…

Strong worst-case performance bounds for episodic reinforcement learning exist but fortunately in practice RL algorithms perform much better than such bounds would predict. Algorithms and theory that provide strong problem-dependent bounds could help illuminate the key features of what makes a RL problem hard and reduc…

This paper improves active learning for Gaussian process regression to handle distributional uncertainty.

Improves policy optimization with polylog(T) regret bounds for stochastic losses.

Optimal strategy identified for minimizing regret in fixed-budget best arm selection.

New approach for pricing evaluation improves on existing methods.

Improved bounds for function approximation in nonlinear sets.

New algorithm expands FTRL framework with improved worst-case regret bounds.