The paper explores the information-theoretic nature of excess risk in machine learning.
problem Understanding the excess risk in machine learning models.
method Formulates the minimax excess risk as a zero-sum game and modifies it to allow swapping of the order of play.
result Proves that under certain conditions, the duality gap is zero, allowing for the application of Bayesian results to provide bounds on minimax excess risk.
Paper improves risk bounds for nonconvex-strongly-concave minimax problems.
problem Achieving sharper risk bounds for nonconvex-strongly-concave minimax problems.
method Using uniform localized convergence to derive high probability generalization error bounds.
result Derives n times faster excess primal risk bounds for popular algorithms.
Paper proposes ZO-SMD for MERO, achieving optimal convergence rates.
problem Minimizing excess risk across all test distributions.
method Zeroth-order stochastic mirror descent algorithm for both smooth and non-smooth MERO.
result Converges at optimal rates of O ( 1 / t ) \mathcal{O}(1/\sqrt{t}) O ( 1/ t ) for estimates and optimization errors. Privacy-preserving binary classification using locally differential private data.
problem Classifying data while protecting individual privacy.
method Locally differential private mechanism followed by a universally consistent classifier.
result Minimax rates of convergence are slower when using private data.
Deep neural network with l_1-regularization achieves nearly optimal risk bounds.
problem Achieving optimal risk bounds in deep learning.
method Empirical risk minimization with l_1-regularization.
result Adaptively nearly-minimax risk bound across various function classes.
Optimizes differentially private kernel learning with random projection.
problem Privacy-preserving learning algorithms with optimal performance.
method Differentially private kernel ERM algorithm based on random projection in reproducing kernel Hilbert space.
result Achieves minimax-optimal excess risk rates for various loss functions.
The paper provides theoretical guarantees for neural network-based anomaly detection.
problem Theoretical guarantees for unsupervised neural network-based anomaly detection.
method Casting anomaly detection as a binary classification problem, establishing non-asymptotic upper bounds and convergence rates.
result The convergence rate on the excess risk matches the minimax optimal rate.
Paper provides optimal statistical guarantees for adversarial robustness in Gaussian classification.
problem Understanding statistical risks for adversarial robustness in Gaussian classification models.
method Established minimax lower bounds and designed an efficient estimator for excess risk.
result Optimal minimax guarantees for excess risk under Gaussian mixture model with AdvSNR.
Last SGD iterate bounds for overparameterized linear regression.
problem Analyzing the last iterate risk bounds of SGD with decaying stepsize for overparameterized linear regression.
method Problem-dependent analysis of last iterate risk bounds of SGD with geometrically decaying stepsize.
result Proved nearly matching upper and lower bounds on the excess risk for last iterate SGD with geometrically decaying stepsize.
Paper tackles linear models with missing values, achieving minimax optimal results.
problem Missing values in real-world data complicate linear model learning.
method Proposes a rigorous setting and a new algorithm leveraging missing data distribution.
result Derives minimax optimal adaptive risk bounds for predictions with missing values.
Study on kernel regression risk in high dimensions using Pinsker bound.
problem Kernel regression risk in high-dimensional inner product spaces.
method Investigation of Pinsker bound for kernel regression on sphere S d \mathbb{S}^{d} S d with sample size n = α d γ ( 1 + o d ( 1 ) ) n = αd^γ(1+o_{d}(1)) n = α d γ ( 1 + o d ( 1 )) . result Exact minimax risk and Pinsker constant identified for kernel regression.
The paper analyzes how optimization algorithms affect the generalization of minimax models.
problem The generalization performance of minimax models trained with different optimization algorithms.
method Analysis of gradient descent ascent (GDA) and proximal point method (PPM) algorithms under convex concave and non-convex non-concave settings.
result The PPM algorithm ensures a bounded excess risk in convex concave problems, while GDA's generalization depends on solving subproblems simultaneously.
Paper introduces robust kernel ridge regression using Cauchy loss for handling various noise types.
problem Developing robust regression methods for noisy data.
method Introduces kernel Cauchy ridge regressor (KCRR) using Cauchy loss function.
result Establishes almost minimax-optimal convergence rate for KCRR in terms of L 2 L_2 L 2 -risk. New research shows that binary classification can be done with noisy data, but only if there are clean samples available.
problem Learning binary classification with instance and label dependent label noise.
method Theoretical analysis and empirical risk minimization.
result Empirical risk minimization achieves the optimal excess risk bound without additional assumptions.
Study on optimal rate of kernel regression for large-dimensional data.
problem Characterizing the upper and lower bounds of kernel regression for large-dimensional data.
method Using Mendelson complexity and metric entropy, the study characterizes the upper and lower bounds of kernel regression for large-dimensional data.
result The minimax rate of the excess risk of kernel regression is \( n^{-1/2} \) for \( n \asymp d^γ \) with \( γ=2, 4, 6, 8, \cdots \).
Develops a robust learning method for unknown context distributions.
problem Learning from data in different, unknown contexts.
method Focuses on excess risks, constructs distribution sets with statistical coverage.
result Shows robustness in worst-case scenarios without sacrificing nominal performance.
This paper analyzes neural network classifiers' performance in binary classification.
problem Performance of neural network classifiers in binary classification problems.
method Plug-in classifiers based on neural networks, considering a more general function class and surrogate loss.
result Dimension-free, uniform rate of convergence for the excess risk of neural networks, showing minimax optimality.
We consider a standard binary classification problem. The performance of any binary classifier based on the training data is characterized by the excess risk. We study Bahadur's type exponential bounds on the minimax accuracy confidence function based on the excess risk. We study how this quantity depends on the comple…
Unified framework for fair classification with group-blindness/awareness guarantees.
problem Challenges in enforcing fairness and group-blindness in binary classification.
method Unified framework based on post-processing procedure, applicable to various group fairness notions.
result Minimax rate-optimality of the proposed algorithm with controlled excess risk.
Two-layer ReLU networks outperform kernel methods in teacher-student settings.
problem Understanding the excess risk of two-layer ReLU neural networks in teacher-student models.
method Investigated a two-phase training process for a student network, comparing it to kernel methods.
result The student network reaches near-global optimality and outperforms kernel methods in minimax optimal rate.
Exact minimax risk derived for linear prediction with sample covariance analysis.
problem Understanding the minimax risk in linear prediction under various covariate distributions.
method Exact minimax risk analysis, leveraging statistical leverage scores and PAC-Bayes techniques.
result The minimax risk is of order d / ( n − d + 1 ) d/(n-d+1) d / ( n − d + 1 ) for any covariate distribution, nearly matching the risk for Gaussian design. SATL adapts to varying smoothness in hypothesis transfer learning.
problem Fixed kernel regularization fails in varying smoothness settings.
method Proposes SATL, a two-phase KRR algorithm with adaptive Gaussian kernels.
result SATL achieves minimax optimality with matching upper and lower bounds.
Paper derives convergence rates for NPMLE in Hellinger distance using deep neural networks.
problem Difficulty in proving convergence of excess risk in nonparametric logistic regression.
method Unified approach for analyzing NPMLE, deriving convergence rates in Hellinger distance.
result Derives nearly optimal convergence rates for NPMLE with deep neural networks.
This paper investigates WDRO for nonparametric regression, achieving robustness against distributional uncertainty.
problem Addressing model misspecification in nonparametric regression under distributional uncertainty.
method Wasserstein distributionally robust optimization (WDRO) with structural distinction based on Wasserstein distance order.
result Achieves a convergence rate of n − 2 β / ( d + 2 β ) n^{-2β/(d+2β)} n − 2 β / ( d + 2 β ) up to logarithmic factors, showing minimax optimality. A central result in statistical theory is Pinsker's theorem, which characterizes the minimax rate in the normal means model of nonparametric estimation. In this paper, we present an extension to Pinsker's theorem where estimation is carried out under storage or communication constraints. In particular, we place limits …
The paper analyzes kernel classifiers' performance in Sobolev spaces and proves their optimality.
problem Theoretical analysis of kernel classifiers' performance in Sobolev spaces.
method Deriving upper and lower bounds on classification excess risk using kernel regression theory and estimating interpolation smoothness.
result The proposed kernel classifier is optimal in Sobolev spaces, with theoretical bounds confirmed by real data.
Gradient descent methods for deep ReLU networks achieve optimal generalization rates.
problem Generalization of gradient descent methods for deep neural networks
method Establishing minimax-optimal rates for GD and SGD with deep ReLU networks
result Gradient descent methods for deep ReLU networks achieve optimal generalization rates
Paper analyzes sparse aggregation in GLMs with Kullback-Leibler risk bounds.
problem Sparse aggregation in GLMs for parameter approximation.
method Exponential weighted aggregation scheme with Kullback-Leibler risk bounds.
result Sharp oracle inequality for Kullback-Leibler risk with leading constant 1 and minimax-optimal rate of aggregation.
Upper bounds and lower bounds show ERM outperforms DG methods in various settings.
problem Limitations of domain generalisation methods in various settings.
method Upper bounds and lower bounds on excess risk of ERM, and analysis of DG settings.
result It is not possible to significantly outperform ERM in DG settings.
We formulate the notion of minimax estimation under storage or communication constraints, and prove an extension to Pinsker's theorem for nonparametric estimation over Sobolev ellipsoids. Placing limits on the number of bits used to encode any estimator, we give tight lower and upper bounds on the excess risk due to qu…
Paper analyzes minimax risks of personalized federated learning algorithms.
problem Statistical heterogeneity among clients in federated learning.
method Minimax analysis of FedAvg and local training approaches.
result Threshold for optimality between FedAvg and local training depends on data heterogeneity.
The paper analyzes risk bounds and Rademacher complexity in batch RL.
problem Estimating/minimizing Bellman error with general value function approximation.
method Characterizes generalization performance using Rademacher complexities of function classes.
result Risk bounds and Rademacher complexities provide insights into batch RL.
Improved DP SO with large Lipschitz parameters, handling outliers and heavy-tailed data.
problem Differential privacy in stochastic optimization with large Lipschitz parameters.
method Assumes bounded k-th order moments, provides linear-time algorithms for smooth convex and non-smooth convex losses.
result Improved risk bounds scaling with k-th moment, not uniform Lipschitz parameter.
Introduces relative information gain for improving Gaussian process regression rates.
problem Improving the sample complexity of estimating or maximizing unknown functions.
method Introduces relative information gain, interpolates between effective dimension and information gain, and proves PAC-Bayesian bounds.
result Obtains minimax-optimal rates of convergence through the relative information gain.
Paper analyzes risk bounds for in-context learning in multiclass classification.
problem Risk bounds for in-context learning in multiclass classification.
method Formalizes tasks as sequences of labeled examples and queries, estimates conditional class probabilities, establishes oracle inequality for KL divergence.
result ICL achieves minimax optimal rate for conditional probability estimation.
Deep learning with noisy gradient descent outperforms linear estimators in high dimensions.
problem Theoretical explanation of deep learning's superiority over linear methods.
method Theoretical analysis of excess risk of a deep learning estimator trained by noisy gradient descent.
result Deep learning achieves a faster learning rate than linear estimators, especially in high dimensions.
New framework analyzes deep learning optimization with finite width networks, revealing generalization gaps and excess risks.
problem Analyzing generalization error of deep learning with finite width networks.
method Formulating neural network training as transportation map estimation and analyzing via infinite dimensional Langevin dynamics.
result Achieves fast learning rate and minimax optimal rates for classification and regression problems.
Optimizes privacy-preserving optimization for heavy-tailed data.
problem Privacy-preserving optimization with heavy-tailed gradients.
method Pure ε-differential privacy framework for Lipschitz extensions.
result Minimax optimal excess-risk rate for pure ε-DP heavy-tailed SCO.
Proposes RaT to mitigate bias in student-teacher estimation.
problem Systematic bias in teacher's predictions propagates to student model.
method Uses teacher to estimate residuals in student's predictions.
result RaT method reduces teacher bias effect and achieves optimal rate.
New method TLC improves transductive learning bounds.
problem Sharp generalization bounds for transductive learning.
method Transductive Local Complexity (TLC) framework.
result Nearly sharp bounds consistent with inductive results.
We consider the problem of binary classification with abstention in the relatively less studied \emph{bounded-rate} setting. We begin by obtaining a characterization of the Bayes optimal classifier for an arbitrary input-label distribution P X Y P_{XY} P X Y . Our result generalizes and provides an alternative proof for the resul…
We present a novel notion of complexity that interpolates between and generalizes some classic existing complexity notions in learning theory: for estimators like empirical risk minimization (ERM) with arbitrary bounded losses, it is upper bounded in terms of data-independent Rademacher complexity; for generalized Baye…
GD outperforms ridge regression and SGD in linear regression problems.
problem Comparing the risks of GD, ridge regression, and SGD in linear regression problems.
method Instance-wise finite-sample risk analysis of GD, ridge regression, and SGD.
result GD outperforms ridge regression and is incomparable with SGD in some cases.
Develops a deep learning framework for various data types.
problem Handling nonparametric regression and classification across different data types.
method Introduces a general framework with two estimators: NPDNN and SPDNN, based on data satisfying generalized Bernstein-type inequalities.
result Both NPDNN and SPDNN estimators are minimax optimal in many classical settings.
The paper analyzes prediction error in nonstationary settings using weighted risk minimization.
problem Prediction under distribution drift and nonstationary conditions.
method General decomposition of excess risk into learning and drift terms, proving oracle inequalities under mixing conditions.
result Oracle inequalities for the learning error, providing bounds that hold uniformly over arbitrary weight classes.
This work characterizes the benefits of averaging schemes widely used in conjunction with stochastic gradient descent (SGD). In particular, this work provides a sharp analysis of: (1) mini-batching, a method of averaging many samples of a stochastic gradient to both reduce the variance of the stochastic gradient estima…
The study provides theoretical guarantees for the statistical performance of optimal decision trees.
problem Theoretical limits on the statistical performance of globally optimal decision trees.
method Sharp oracle inequalities and uniform concentration framework based on Rademacher complexity.
result Derivation of minimax optimal rates for piecewise sparse heterogeneous anisotropic Besov space.
Two-layer neural networks can overfit without increasing risk when data is noisy.
problem Understanding why neural networks can overfit without increasing risk in noisy data.
method Combining bias and variance analysis in a high-dimensional setting.
result The excess learning risk of the interpolator decays under mild conditions.