Introduces PAC-Bayes bounds for understanding learning procedures.
problem Understanding the generalization ability of learning procedures.
method PAC-Bayesian bounds and their applications to neural networks.
result Simplified version of localization technique described.
Derandomizing PAC-Bayes bounds for smooth loss functions
problem Derandomizing PAC-Bayes bounds for smooth loss functions
method Exploiting smoothness properties of both the loss and the predictor class
result Bounds for deterministic predictors that involve flatness quantities
PAC-Bayes bound requires prior to place mass on high-performing predictors.
problem Explaining generalization in machine learning.
method Analyzing necessary conditions for PAC-Bayes bounds to provide meaningful generalization guarantees.
result Achieving a target generalisation level requires the prior to place sufficient mass on high-performing predictors.
In spite of several notable efforts, explaining the generalization of deterministic non-smooth deep nets, e.g., ReLU-nets, has remained challenging. Existing approaches for deterministic non-smooth deep nets typically need to bound the Lipschitz constant of such deep nets but such bounds are quite large, may even incre…
Derives bounds for deterministic predictors using smooth loss functions.
problem Generalizing probabilistic predictors to deterministic ones.
method Exploits smoothness properties of loss and predictor classes, controlling the Jensen gap class through Rademacher complexity.
result Derives bounds for deterministic predictors involving flatness quantities from Jacobians and Hessians.
Bayes predictor remains robust to ignorable missingness shifts.
problem Challenges in prediction with missing covariates and shifts in missingness reasons.
method Bayesian approach and different prediction methods.
result Bayes predictor remains unchanged by ignorable shifts, but robust prediction requires disregarding missingness for non-ignorable shifts.
The paper improves PAC-Bayes bounds for data-dependent predictors.
problem Guaranteeing the quality of predictions on unseen examples.
method Basic PAC-Bayes inequality for stochastic kernels, leading to various bounds.
result Validates PAC-Bayes bounds without fixed 'data-free' priors and bounded losses.
New method uses quotient predictor space for better PAC-Bayes bounds, reducing KL divergence and improving model performance.
problem Overparameterized models with continuous symmetries can lead to biased predictions.
method Perform PAC-Bayesian analysis on quotient predictor space, constructing a canonical prior that reflects model's implicit bias.
result The new prior reduces KL divergence and improves model performance in experiments.
This work uses PAC-Bayes for structured prediction with ILE, yielding insights and algorithms.
problem Structured prediction with interdependent outputs and implicit loss embeddings.
method PAC-Bayes perspective applied to ILE framework, deriving generalization bounds and learning algorithms.
result Two learning algorithms derived from PAC-Bayes bounds, analyzed and implemented.
New method detects information leakage using approximate Bayes predictor.
problem Unintentional exposure of sensitive information via observable data.
method Statistical learning theory and information theory framework, approximating Bayes predictor's log-loss and accuracy.
result MI can be accurately estimated to detect ILs, outperforming state-of-the-art baselines.
Paper presents new training methods for neural networks with tighter risk certificates.
problem Training probabilistic neural networks with tighter risk certificates.
method Derived from PAC-Bayes bounds, two training objectives implemented for the first time in neural networks.
result Competitive test set errors and non-vacuous risk bounds with tighter values than previous results.
Bayesian framework reduces online optimization regret.
problem Sequential optimization in dynamic environments with bounded losses.
method PAC-Bayes theory and Bayesian updating principles.
result Achieves O ( T ) \mathcal{O}(\sqrt{T}) O ( T ) regret for bounded losses. New methods for CI testing under model misspecification.
problem Challenges in CI testing with misspecified models.
method Proposes new approximations and upper bounds for testing errors of regression-based CI tests.
result Introduces the Rao-Blackwellized Predictor Test (RBPT) robust against misspecified inductive biases.
Memory-based models can learn to approximate Bayes-optimal predictors for non-stationary data.
problem Learning from non-stationary data with unobserved switching points.
method Memory-based neural models, including Transformers, LSTMs, and RNNs, trained to minimize log loss.
result Memory-based models can accurately approximate known Bayes-optimal algorithms and perform Bayesian inference over latent switching points.
The problem of forecasting conditional probabilities of the next event given the past is considered in a general probabilistic setting. Given an arbitrary (large, uncountable) set C of predictors, we would like to construct a single predictor that performs asymptotically as well as the best predictor in C, on any data.…
We consider using an ensemble of binary classifiers for transductive prediction, when unlabeled test data are known in advance. We derive minimax optimal rules for confidence-rated prediction in this setting. By using PAC-Bayes analysis on these rules, we obtain data-dependent performance guarantees without distributio…
We explore the family of methods "PAC-Bayes with Backprop" (PBB) to train probabilistic neural networks by minimizing PAC-Bayes bounds. We present two training objectives, one derived from a previously known PAC-Bayes bound, and a second one derived from a novel PAC-Bayes bound. Both training objectives are evaluated o…
Smart Bayes integrates generative and discriminative features for improved classification.
problem Improving classification performance by combining generative and discriminative modeling.
method Integrates generative likelihood-ratio features into a logistic-regression-style classifier.
result Often outperforms logistic regression and Naive Bayes in simulations and real data.
New bounds link flat minima to good generalisation in overparameterized models.
problem Understanding the relationship between flat minima and generalisation in overparameterized machine learning models.
method Combining PAC-Bayes, Poincaré, and Log-Sobolev inequalities to derive generalisation bounds involving gradient terms.
result Flat minima positively influence generalisation performance, highlighting the benefits of the optimisation phase.
Enhances predictive models against misspecification and outliers.
problem Suboptimal generalization under misspecification and outliers.
method Combines PAC m ^m m ensemble bounds with a generalized logarithm score function. result Produces predictive distributions resistant to both misspecification and outliers.
We propose a penalized orthogonal-components regression (POCRE) for large p small n data. Orthogonal components are sequentially constructed to maximize, upon standardization, their correlation to the response residuals. A new penalization framework, implemented via empirical Bayes thresholding, is presented to effecti…
PROBE algorithm efficiently solves sparse high-dimensional linear regression.
problem Sparse high-dimensional linear regression models with complex parameter spaces.
method Partitioned empirical Bayes ECM algorithm for computationally efficient MAP estimation.
result PROBE algorithm provides robust and efficient coordinate-wise optimization.
This research improves binary classification by balancing overfitting and generalization with a novel Bayesian approach.
problem Improving binary classification models to avoid overfitting and generalize well.
method Introduces a PAC-Bayes type learning rule with a balancing parameter λ to balance training error and KL divergence to a prior.
result A choice of λ ensures uniformly vanishing excess loss, even in the agnostic case, by under-regularizing or over-regularizing appropriately.
New method certifies deep graph classifiers with tighter risk bounds.
problem Certifying the reliability of deep graph classifiers.
method Linearized deep assignment flows with random initial conditions, using PAC-Bayes risk certification.
result Computes tighter out-of-sample risk certificates efficiently.
Transformer attention layers solve single-location regression tasks.
problem Understanding token-wise sparsity and internal linear representations in attention-based models.
method Introduce single-location regression task and a simplified predictor based on self-attention layers.
result Transformer attention layers are asymptotically Bayes optimal and can learn underlying structures effectively.
New PAC-Bayes bounds for unbounded loss functions.
problem Generalization bounds for learning problems with unbounded loss functions.
method Introducing HYPE, a new notion for loss range, and deriving a novel PAC-Bayesian generalization bound.
result PAC-Bayes framework extended to unbounded loss functions.
New bounds prevent degradation in high-dimensional signal estimation.
problem Statistical learning bounds degradation with increasing dimensionality.
method Investigates linear prediction rules under structural assumptions.
result Derives upper and lower bounds on generalization error.
The paper proposes multicalibration to improve matching in graphs with imperfect predictors.
problem Finding the best matching in graphs with imperfect predictors.
method Introduces multicalibration as a fairness notion to ensure unbiasedness on protected sets of contexts.
result Constructing a multicalibrated predictor that outperforms standard optimal rules in matching algorithms.
Bayesian Hierarchical Invariant Prediction refines ICP for better scalability and prior integration.
problem Improving computational scalability and invariance testing for causal inference.
method Bayesian Hierarchical structure to test invariance under heterogeneous data.
result Demonstrated improved scalability and potential as an alternative to ICP.
Paper proves fair classification can be done via simple thresholding.
problem Achieving fair binary classification subject to group fairness constraints.
method Proves Bayes optimal fair learning rule is a group-wise thresholding rule over the Bayes regressor with randomization.
result Proposes an efficient unconstrained optimization algorithm for post-processing fair classification.
NeuMiss networks tackle supervised learning with missing values, offering efficient and robust predictions.
problem Challenges in supervised learning with missing values, especially when the response is a linear function of the complete data.
method Derive analytical form of optimal predictor under linearity assumption and various missing data mechanisms. Propose NeuMiss networks using multiplication by missingness indicator.
result Upper bound on Bayes risk and good predictive accuracy with independent complexity of missing data patterns.
It has been argued that in supervised classification tasks, in practice it may be more sensible to perform model selection with respect to some more focused model selection score, like the supervised (conditional) marginal likelihood, than with respect to the standard marginal likelihood criterion. However, for most Ba…
Bayesian approach optimizes in-context learning for state space models.
problem Optimizing in-context learning for state space models.
method Bayesian optimal sequential prediction over latent sequence tasks.
result Bayesian optimal predictor converges to posterior predictive mean.
The paper proves how Transformers learn from context and generalize well.
problem Understanding how Transformers generalize from diverse tasks.
method Developed a statistical theory for in-context learning, separating risk into Bayes Gap and Posterior Variance.
result The Posterior Variance is task-independent, and the Bayes Gap decreases with more in-context examples.
Optimizes pruning masks for neural networks using probabilistic fine-tuning and PAC-Bayes bounds.
problem Improving neural network performance through adaptive pruning of weights.
method Optimizes stochastic pruning masks by minimizing expected loss, considering data-adaptive regularization and feature alignment.
result Probabilistic fine-tuning leads to improved test error over baseline methods in neural networks.
Attention-only transformers learn from context via two stages of inference.
problem Learning from corrupted token sequences in minimal transformers.
method Two-stage empirical Bayes interpretation: kernel-weighted posterior mean and particle dynamics.
result Effective denoising without explicit noise schedules, showing posterior-mean recovery under asymptotic conditions.
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.
In the era of "big data", it is becoming more of a challenge to not only build state-of-the-art predictive models, but also gain an understanding of what's really going on in the data. For example, it is often of interest to know which, if any, of the predictors in a fitted model are relatively influential on the predi…
New framework shows cross-attention improves multi-modal in-context learning.
problem Understanding multi-modal in-context learning in neural networks.
method Mathematical framework and linearized cross-attention mechanism.
result Cross-attention mechanism is provably optimal for multi-modal in-context learning.
Survey on calibration in machine learning, viewing it as indistinguishability.
problem Evaluating continuous probability predictions in discrete outcome settings.
method Defining and measuring calibration error through indistinguishability.
result Calibration measures quantify distinguishability between hypothesized and real-world outcomes.
Optimal classifiers derived from GMMs are approximated by deep neural networks.
problem Binary classification of high-dimensional overlapping Gaussian mixtures.
method Closed-form expressions for Bayes optimal decision boundaries derived from GMMs' eigenstructure. Empirical validation through synthetic and real-world data.
result Deep neural networks approximate optimal classifiers for GMMs, with decision thresholds related to covariance eigenvectors.
Continuum transformers learn operators in context via gradient descent.
problem Generalizing transformers to handle infinite-dimensional inputs for in-context learning.
method Gradient descent in an operator RKHS, leveraging generalized representer theorems and gradient flows.
result Operator learned in context is Bayes Optimal Predictor in infinite depth limit.
New method for spatiotemporal data regression using Gaussian processes.
problem Regression in spatiotemporal random fields.
method Empirical Bayes approach, tight Gaussian measures, truncation scheme.
result Effective dimension reduction through time-varying angular spectra.
New framework improves adversarial robustness in one-stage L2D.
problem Adversarial robustness in one-stage Learning-to-Defer (L2D).
method Formalizes attacks, proposes cost-sensitive adversarial surrogate losses, establishes theoretical guarantees.
result Improves robustness against untargeted and targeted attacks while preserving clean performance.
A new method predicts true classes from positive and unlabeled data with additional labeled observations.
problem Predicting true classes from positive and unlabeled data with selection bias.
method Introduces augmented PU prediction, allowing feature-dependent labeling, and compares various empirical Bayes rules.
result The variational autoencoder-based method performs similarly or better than other methods and improves accuracy for unlabeled samples.
Gradient-based optimization improves variational empirical Bayes regression.
problem Sparse, large-scale multiple regression models.
method Gradient-based optimization (GradVI) for variational empirical Bayes (VEB) regression.
result GradVI produces similar predictive performance to CAVI but converges faster and is faster in certain settings.
Paper introduces a method to create robust representations against covariate shifts.
problem Distribution shift between training and testing data in machine learning.
method Introduces a variational objective with two components: discriminative representation and invariant support.
result Optimal representations ensure robustness to covariate shifts, improving performance on DomainBed.
Modern statistical applications involving large data sets have focused attention on statistical methodologies which are both efficient computationally and able to deal with the screening of large numbers of different candidate models. Here we consider computationally efficient variational Bayes approaches to inference …