We show how to compute the Bayes error-rate for speaker verifiers.
problem How many errors does a speaker verifier make in a hundred trials?
method We compute the Bayes error-rate using calibrated likelihood ratios and user-supplied prior probabilities.
result The Bayes error-rate is upper bounded by the minimum of EER, P, and 1-P.
We address the problem of learning to benchmark the best achievable classifier performance. In this problem the objective is to establish statistically consistent estimates of the Bayes misclassification error rate without having to learn a Bayes-optimal classifier. Our learning to benchmark framework improves on previ…
This work bounds classification error in machine learning for low Bayes error conditions.
problem Understanding the error mismatch between Bayes error and model-based classification error.
method Applying classification error bounds to study the relationship with Kullback-Leibler divergence and proposing a linear approximation for low Bayes error conditions.
result A linear approximation of the classification error bound for low Bayes error conditions is proposed.
Meta learning of optimal classifier error rates allows an experimenter to empirically estimate the intrinsic ability of any estimator to discriminate between two populations, circumventing the difficult problem of estimating the optimal Bayes classifier. To this end we propose a weighted nearest neighbor (WNN) graph es…
New PAC-Bayes bound controls multiple error types simultaneously.
problem Current PAC-Bayes bounds are limited to scalar metrics.
method Bounding KL divergence between empirical and true probabilities of multiple error types.
result First PAC-Bayes bound for rich information-rich certificates.
Study shows how classifiers can approach Bayes error in high-dimensional settings.
problem Generalization error in high-dimensional perceptrons.
method Proved a formula for generalization error using convex optimization and observed that logistic and hinge regression can approach Bayes error closely.
result Logistic and hinge regression can approach Bayes-optimal generalization error closely in high-dimensional settings.
Unified derivation of PAC-Bayes and MI bounds for general VC classes with fast rates.
problem Generalization bounds for machine learning models with VC classes.
method Unified derivation of conditional PAC-Bayesian and mutual information bounds, including MAC-Bayesian bounds.
result Nontrivial bounds for general VC classes and faster rates for specific conditions.
We propose a new splitting criterion for a meta-learning approach to multiclass classifier design that adaptively merges the classes into a tree-structured hierarchy of increasingly difficult binary classification problems. The classification tree is constructed from empirical estimates of the Henze-Penrose bounds on t…
Lower bounds on Bayes risk for realizable models derived using information theory.
problem Deriving lower bounds on Bayes risk for realizable machine learning models.
method Information-theoretic analysis using rate-distortion theory and mutual information.
result Lower bounds on Bayes risk for realizable models, matching known bounds up to logarithmic factors.
Bayes Error Rate estimators are evaluated for accuracy and sample requirements.
problem Evaluating the accuracy and sample requirements of Bayes Error Rate estimators.
method Monte Carlo simulations with synthetic data and real-world scenarios.
result k-Nearest Neighbor (kNN) is the most accurate non-parametric estimator.
We study the statistical performance of semidefinite programming (SDP) relaxations for clustering under random graph models. Under the Z2 Synchronization model, Censored Block Model and Stochastic Block Model, we show that SDP achieves an error rate of the form \[ \exp\Big[-\big(1-o(1)\big)\bar{n} I^* \Bi…
In many signal detection and classification problems, we have knowledge of the distribution under each hypothesis, but not the prior probabilities. This paper is aimed at providing theory to quantify the performance of detection via estimating prior probabilities from either labeled or unlabeled training data. The erro…
Improved Naive Bayes classifier with neural network models.
problem Limited complexity handling and independence assumption in Naive Bayes.
method Introducing Neural Naive Bayes and Neural Pooled Markov Chain models.
result Error rate reduced by 4.5 on IMDB dataset.
A new method estimates Bayes error for deep networks, suggesting they may have reached the limit.
problem Evaluating the performance of deep learning models and detecting overfitting.
method A simple and direct Bayes error estimator based on uncertainty of class assignments.
result Deep networks may have reached the Bayes error limit for benchmark datasets.
Empirical Bayes rates via variational approximations and prior decomposition.
problem Nonparametric and high-dimensional inference convergence rates.
method Variational perspective and prior decomposition.
result Empirical Bayes posterior rates derived from variational Bayes.
Paper estimates optimal classification error with soft labels and calibration.
problem Estimating the optimal classification error with soft labels and calibration.
method Extends previous work on soft labels to estimate Bayes error, addressing bias and corrupted labels.
result The method provides a statistically consistent estimator of the Bayes error, even with imperfectly calibrated soft labels.
Researchers compute Bayes error for classification models using normalizing flows.
problem Evaluating the inherent difficulty of classification problems.
method Invertible transformations and Gaussian base distributions to compute Bayes error.
result State-of-the-art models can achieve near-optimal accuracy but not always.
New research shows existing information-theoretic methods can't establish minimax rates for gradient descent in stochastic convex optimization.
problem Establishing minimax rates for gradient descent in stochastic convex optimization using information-theoretic methods.
method Examined several information-theoretic frameworks including input-output mutual information bounds, conditional mutual information bounds, PAC-Bayes bounds, and their variants.
result Proved that none of the examined information-theoretic frameworks can establish minimax rates for gradient descent in stochastic convex optimization.
Minimizes indecisions in selective classification to control misclassification rates.
problem Controlling misclassification rates in high-risk scenarios.
method Using indecisions to control misclassification rates, even below Bayes optimal.
result Control of misclassification rates to any user-specified level, even below Bayes optimal.
Paper introduces new bounds linking data compressibility to generalization error.
problem Establishing data-dependent generalization bounds.
method Variable-size compressibility framework linking generalization error to compression rate of input data.
result New bounds depend on empirical data measure, subsuming existing PAC-Bayes and intrinsic dimension bounds.
PrAda-GAN improves synthetic data generation under differential privacy.
problem Generating synthetic data under differential privacy with marginal-based methods.
method Sequential generator architecture integrating GAN and marginal-based approaches, with adaptive regularization of Bayes network structure.
result PrAda-GAN outperforms existing methods in privacy-utility trade-off on synthetic and real-world datasets.
Paper proposes a method to estimate total variation distance for synthetic data fidelity.
problem Assessing the fidelity of synthetic data generated by AI.
method Discriminative approach to estimate total variation distance between two distributions.
result Estimation of total variation distance reduces to quantifying Bayes risk in classification.
Study shows multi-distribution learning has slower rates than single-task learning.
problem Understanding the statistical complexity of learning from heterogeneous sources.
method Structured hypothesis-testing framework to capture the statistical cost of certifying near-optimality under bounded noise.
result Learning across multiple distributions incurs slow rates scaling with k/ε2, even under constant noise levels. Paper improves PAC-Bayes bounds using a better-than-KL divergence.
problem Estimating the generalization error of stochastic algorithms.
method Developed new PAC-Bayes bounds with a novel divergence.
result Achieved strictly tighter bounds than the KL divergence.
The developments of Rademacher complexity and PAC-Bayesian theory have been largely independent. One exception is the PAC-Bayes theorem of Kakade, Sridharan, and Tewari (2008), which is established via Rademacher complexity theory by viewing Gibbs classifiers as linear operators. The goal of this paper is to extend thi…
We study the rates of convergence from empirical surrogate risk minimizers to the Bayes optimal classifier. Specifically, we introduce the notion of \emph{consistency intensity} to characterize a surrogate loss function and exploit this notion to obtain the rate of convergence from an empirical surrogate risk minimizer…
We show how to compute lower bounds for the supremum Bayes error if the class-conditional distributions must satisfy moment constraints, where the supremum is with respect to the unknown class-conditional distributions. Our approach makes use of Curto and Fialkow's solutions for the truncated moment problem. The lower …
Develops a fast variational approximation for high-dimensional empirical Bayes posteriors.
problem Optimal posterior computation in high-dimensional settings with prior tails effect.
method Variational approximation of empirical Bayes posterior with data-driven centers and thin-tailed conjugate priors.
result Retains optimal concentration rate properties and superior performance compared to existing methods.
Calibration without labels in multiple testing
problem Interpretable error probabilities in large-scale hypothesis testing
method Constructing pseudo-labels from spacings of ordered p-values result Finding that q-value can be severely miscalibrated The paper improves PAC-Bayes bounds for losses with finite moments.
problem Bounding generalization for losses with heavy tails and finite moments.
method Truncation method and PAC-Bayes bounds for unbounded losses with heavy tails and bounded variance.
result Bounds interpolate between slow and fast rates depending on the moment.
Bayes-optimal learning of deep random networks with Gaussian weights is studied.
problem Learning a target function corresponding to a deep, extensive-width, non-linear neural network with random Gaussian weights.
method Closed-form expressions for Bayes-optimal test error, ridge regression, kernel and random features regression are computed.
result Optimally regularized ridge regression and kernel regression achieve Bayes-optimal performances, while logistic loss yields a near-optimal test error for classification.
Bayes classifier cannot be learned from noisy labels without knowing noise distribution.
problem Learning a Bayes classifier from noisy labels when the noise distribution is unknown.
method Demonstrates the identifiability issues and proposes a simple algorithm for learning the Bayes decision rule.
result The Bayes decision rule is generally unidentified and cannot be learned without knowing the noise distribution.
Research on nearest-neighbor methods tends to focus somewhat dichotomously either on the statistical or the computational aspects -- either on, say, Bayes consistency and rates of convergence or on techniques for speeding up the proximity search. This paper aims at bridging these realms: to reap the advantages of fast …
Paper introduces ILD algorithm to determine Bayes error for binary classification.
problem Determining the best possible performance in binary classification problems.
method Model-agnostic ILD algorithm to calculate Bayes error.
result Provides intrinsic limits of any binary classification algorithm.
We study high-dimensional asymptotic performance limits of binary supervised classification problems where the class conditional densities are Gaussian with unknown means and covariances and the number of signal dimensions scales faster than the number of labeled training samples. We show that the Bayes error, namely t…
New method compresses large sample data for faster discriminant analysis.
problem Large sample sizes in discriminant analysis increase computational burden.
method Proposes a new compression approach for reducing training samples.
result Significant computational gains and superior predictive ability compared to random sub-sampling.
New work shows limits of certifying neural network robustness.
problem Certified training improves robustness but decreases accuracy.
method Bayes error analysis to investigate robustness limits.
result Upper bound for certified robust accuracy established.
The study analyzes multi-class teacher-student perceptron performance and generalization errors.
problem Analyzing multi-class classification with the teacher-student perceptron.
method Deriving asymptotic expressions for Bayes-optimal and empirical risk minimization (ERM) generalization errors.
result Regularised cross-entropy minimization yields close-to-optimal accuracy for multi-class classification.
Study shows MDA's effectiveness even when more components are assumed than in actual data.
problem Classification error in overspecified Mixture Discriminant Analysis.
method Two-component Gaussian mixture model, EM algorithm, theoretical analysis of convergence and error rates.
result EM algorithm converges exponentially fast to Bayes risk with suitable initialization.
Study improves theoretical understanding of Bayesian deep learning for classification tasks.
problem Theoretical gap in understanding Bayesian approaches in deep learning for classification.
method PAC-Bayes bounds techniques and Spike-and-Slab priors for sparse deep learning.
result Established non-asymptotic results for prediction error, achieving minimax optimal rates.
Sequence-to-sequence models, such as attention-based models in automatic speech recognition (ASR), are typically trained to optimize the cross-entropy criterion which corresponds to improving the log-likelihood of the data. However, system performance is usually measured in terms of word error rate (WER), not log-likel…
We propose an empirical Bayes estimator based on Dirichlet process mixture model for estimating the sparse normalized mean difference, which could be directly applied to the high dimensional linear classification. In theory, we build a bridge to connect the estimation error of the mean difference and the misclassificat…
Paper estimates FPR of Bayes classifier using soft labels.
problem Determining optimal classifier performance.
method Uses soft labels and denoising technique.
result Consistent and unbiased FPR estimator developed.
The purpose of this paper is to propose a time-varying vector autoregressive model (TV-VAR) for forecasting multivariate time series. The model is casted into a state-space form that allows flexible description and analysis. The volatility covariance matrix of the time series is modelled via inverted Wishart and singul…
New PAC-Bayes training method improves model generalization for unbounded loss.
problem Improving generalization of complex models under unbounded loss.
method Established new PAC-Bayes bound for unbounded loss, jointly training prior and posterior.
result Outperforms existing PAC-Bayes training algorithms and matches ERM accuracy.
Study shows more frequent communication in FL reduces model's generalization power.
problem Understanding how often clients should communicate with a server in federated learning.
method PAC-Bayes and rate-distortion theoretic bounds on generalization error.
result More frequent communication with the server decreases model's generalization power.
In this work, we propose a PAC-Bayes bound for the generalization risk of the Gibbs classifier in the multi-class classification framework. The novelty of our work is the critical use of the confusion matrix of a classifier as an error measure; this puts our contribution in the line of work aiming at dealing with perfo…
Bayes-optimal learning of a neural network with quadratic activations is achieved with GAMP-RIE.
problem Learning a neural network with quadratic activations from quadratic samples.
method Combining approximate message passing with rotationally invariant matrix denoising.
result Derives a closed-form expression for Bayes-optimal test error.