Cross-regularization adapts model complexity during training.
problem Manual tuning of model complexity for overfitting prevention.
method Directly adapts regularization parameters through validation gradients during training.
result Organic emergence of architecture-specific regularization during training.
We consider the problem of estimating the parameters of the covariance function of a Gaussian process by cross-validation. We suggest using new cross-validation criteria derived from the literature of scoring rules. We also provide an efficient method for computing the gradient of a cross-validation criterion. To the b…
Analysis of cross-validation for early-stopped gradient descent in high-dimensional regression.
problem Inconsistency of GCV for early-stopped GD in high-dimensional least squares regression.
method Theoretical analysis of GCV and LOOCV applied to early-stopped GD in high-dimensional least squares regression.
result LOOCV converges uniformly to the prediction risk of early-stopped GD, while GCV is generically inconsistent.
Cross-validation is the workhorse of modern applied statistics and machine learning, as it provides a principled framework for selecting the model that maximizes generalization performance. In this paper, we show that the cross-validation risk is differentiable with respect to the hyperparameters and training data for …
ScoreStop uses gradient tests to stop gradient boosting early.
problem Overfitting in gradient boosted decision trees.
method ScoreStop uses a functional score test based on gradients to stop boosting.
result ScoreStop is competitive with loss-based early stopping methods.
Develops anytime-valid stopping rules for SGD based on observed trajectory.
problem Stopping stochastic gradient descent (SGD) based on observed trajectory.
method Develops anytime-valid confidence sequences for stochastic gradient methods.
result Statistically valid, time-uniform stopping rules for SGD across convex and nonconvex settings.
This work examines the convergence of stochastic gradient-based optimization algorithms that use early stopping based on a validation function. The form of early stopping we consider is that optimization terminates when the norm of the gradient of a validation function falls below a threshold. We derive conditions that…
New method improves online nonparametric estimators with minimal extra computation.
problem Model selection and hyperparameter tuning for online nonparametric estimators.
method Weighted rolling validation procedure for online cross-validation.
result Improves base estimators to achieve better heuristic performance and adaptive convergence rate.
In this paper, motivated by the works of Bakry et. al in finding sharp Li-Yau type gradient estimate for positive solutions of the heat equation on complete Riemannian manifolds with nonzero Ricci curvature lower bound, we first introduce a general form of Li-Yau type gradient estimate and show that the validity of suc…
Improved scalable machine learning under heavy-tailed data.
problem Machine learning scalability under heavy-tailed data without strong convexity.
method Simple robust validation sub-routine to boost confidence in gradient-based sub-processes.
result Substantial improvement in dimension dependence without strong convexity.
We present a novel regularization approach to train neural networks that enjoys better generalization and test error than standard stochastic gradient descent. Our approach is based on the principles of cross-validation, where a validation set is used to limit the model overfitting. We formulate such principles as a bi…
Early stopping is a widely used technique to prevent poor generalization performance when training an over-expressive model by means of gradient-based optimization. To find a good point to halt the optimizer, a common practice is to split the dataset into a training and a smaller validation set to obtain an ongoing est…
Tuning hyperparameters of learning algorithms is hard because gradients are usually unavailable. We compute exact gradients of cross-validation performance with respect to all hyperparameters by chaining derivatives backwards through the entire training procedure. These gradients allow us to optimize thousands of hyper…
The paper proposes a gradient-based method for multi-penalty Ridge regression.
problem Optimizing multiple regularization hyperparameters for linear regression.
method Gradient-based optimization through matrix differential calculus.
result The method outperforms traditional regularization techniques like LASSO and Ridge.
We introduce a novel algorithm for the detection of possible sample corruption such as mislabeled samples in a training dataset given a small clean validation set. We use a set of inclusion variables which determine whether or not any element of the noisy training set should be included in the training of a network. We…
agtboost speeds up gradient tree boosting with automatic complexity adjustment.
problem Speeding up and simplifying gradient tree boosting computations.
method Adaptive gradient tree boosting with automatic complexity adjustment and feature importance.
result Significant decrease in computation time and simplification of model complexity.
Local constancy of index for certain gradient mappings proved.
problem Proving the local constancy of the index for specific gradient mappings.
method Using a more general theorem for quasiregular gradient mappings, deducing the result from the Hessian's properties.
result The index is locally constant for C1,1 functions with uniformly positive determinant Hessian almost everywhere. New methods validate a hypothesis explaining how neural nets generalize well.
problem Why over-parameterized nets generalize well despite memorizing training data.
method Developed new algorithms to suppress weak gradient directions without per-example gradients.
result Validated a hypothesis about gradient directions and their role in generalization.
New method optimizes hyperparameters for non-smooth problems efficiently.
problem Efficiently tuning hyperparameters for non-smooth cost functions.
method Combines hyperparameter search with proximal gradient updates.
result Method converges to local optimum of LOO validation error.
Python package for ordinal regression using gradient boosting.
problem Handling ordinal variables in machine learning.
method Gradient boosting with latent variable framework.
result Performs joint optimization of latent function and threshold vector.
Paper proposes HCDC to improve hyperparameter search efficiency.
problem Poor generalizability of dataset condensation across different hyperparameters.
method HCDC algorithm that matches hyperparameter gradients for synthetic validation dataset.
result HCDC effectively maintains validation-performance rankings of models.
Gradient descent memorizes many Gaussians efficiently.
problem Memorizing many Gaussians with minimal parameters.
method Gradient descent on a depth-two neural network.
result One step of gradient descent memorizes $Ω\left(\frac{dq}{\log^4(d)}
ight)$ Gaussians.
In high-dimensional and/or non-parametric regression problems, regularization (or penalization) is used to control model complexity and induce desired structure. Each penalty has a weight parameter that indicates how strongly the structure corresponding to that penalty should be enforced. Typically the parameters are c…
New study shows neural networks can generalize without gradient descent, especially in deep settings.
problem Whether neural networks need gradient descent for generalization.
method Theoretical study of matrix factorization with linear and non-linear activation, comparing gradient descent to Guess & Check.
result Generalization under Guess & Check deteriorates with increasing width but improves with depth, challenging conventional wisdom.
New method corrects bias in feature importance measures of GBM.
problem Bias in feature importance measures of GBM.
method Cross-validated unbiased base learners.
result Significant improvement in feature importance measures with minimal computational cost.
M-FISHER detects and adapts to streaming data shifts with statistical validity and stability.
problem Detecting and adapting to distributional shifts in streaming data.
method Constructs an exponential martingale from non-conformity scores and applies Ville's inequality for detection. Fisher-preconditioned updates for adaptation.
result Establishes M-FISHER as a principled approach for robust, anytime-valid detection and geometrically stable adaptation.
SGLRW improves robustness of stochastic gradient MCMC methods.
problem Sensitivity to minibatch size and gradient noise in stochastic-gradient MCMC methods.
method Proposes Stochastic Gradient Lattice Random Walk (SGLRW) with lattice-based discretization.
result SGLRW remains stable in regimes where SGLD fails, including heavy-tailed gradient noise.
New algorithm helps escape saddle points in optimization problems.
problem Optimizing smooth non-convex functions to avoid saddle points.
method Perturbed Saddle-escape Descent (PSD) algorithm with explicit constants.
result PSD finds approximate second-order stationary points efficiently.
Optimizes SGLD noise structure for better generalization bounds.
problem Improving generalization bounds for large models trained with SGLD.
method Manipulates the noise structure in SGLD to optimize information-theoretical bounds.
result Optimal noise covariance is the square root of the expected gradient covariance under certain constraints.
Gradient descent dynamics in nonconvex models explained with universality.
problem Understanding long-time behavior of nonconvex gradient descent.
method Developed a state evolution system for tracking gradient descent iterates.
result Gradient descent iterates are approximately independent of data and strongly incoherent with feature vectors.
Study tests if input gradients highlight discriminative features, finds they often fail.
problem Validity of assumption that input gradients highlight discriminative features in model predictions.
method Developed DiffROAR framework and BlockMNIST dataset to test assumption on four benchmarks.
result Input gradients of standard models often fail to highlight discriminative features, while robust models do.
This paper explores how train-validation splits help in NAS to prevent overfitting.
problem NAS overfits with train-validation splits and needs better generalization guarantees.
method Established refined properties of validation loss and risk for NAS.
result NAS with train-validation splits can select the most generalizable model.
Label noise may affect the generalization of classifiers, and the effective learning of main patterns from samples with noisy labels is an important challenge. Recent studies have shown that deep neural networks tend to prioritize the learning of simple patterns over the memorization of noise patterns. This suggests a …
Optimal defenses protect FL models from gradient reconstruction attacks.
problem Gradient reconstruction attacks compromise FL models by recovering original data from shared gradients.
method Derive a theoretical lower bound of reconstruction error, customize noise and pruning defenses, and achieve optimal trade-off between leakage and utility.
result Our methods outperform Gradient Noise and Pruning in protecting training data and maintaining model utility.
Superconvergence of differential structure on discretized surfaces is studied in this paper. The newly introduced geometric supercloseness provides us with a fundamental tool to prove the superconvergence of gradient recovery on deviated surfaces. An algorithmic framework for gradient recovery without exact geometric i…
Exchanging gradients is a widely used method in modern multi-node machine learning system (e.g., distributed training, collaborative learning). For a long time, people believed that gradients are safe to share: i.e., the training data will not be leaked by gradient exchange. However, we show that it is possible to obta…
The paper provides gradient estimates for solutions on manifolds with integral Ricci bounds.
problem Global regularity estimates for solutions of Δu=f on Riemannian manifolds. method Proves Lp-gradient estimates under integral Ricci bounds and constructs a counterexample. result Optimal constant lower bounds on Ricci curvature are shown in the pointwise sense.
Optimizes hyperparameter tuning for models using approximate leave-one-out cross-validation.
problem Finding optimal hyperparameters for regularized models using approximate leave-one-out cross-validation.
method Derive efficient formulas for gradient and hessian of approximate leave-one-out cross-validation, apply second-order optimization.
result Demonstrates the effectiveness of the approach on real-world data sets.
New theory validates the use of invariant predictors for OOD generalization.
problem Ensuring predictors generalize well across unseen environments.
method Developed new theoretical conditions and derived an Inter Gradient Alignment algorithm.
result Validated the necessity of invariant predictors for OOD optimality.
Multi-subject fMRI data analysis is an interesting and challenging problem in human brain decoding studies. The inherent anatomical and functional variability across subjects make it necessary to do both anatomical and functional alignment before classification analysis. Besides, when it comes to big data, time complex…
U-Clip clips gradients to reduce bias without discarding them.
problem Reduces bias in stochastic gradient descent.
method Maintains a buffer of clipped gradients to add to next iteration.
result U-Clip updates are unbiased on average.
The paper validates a method for recovering over-parameterized matrices and images from noisy measurements.
problem Recovering a low-rank matrix from noisy measurements when the rank is unknown.
method Using gradient descent with small random initialization on a nonconvex objective function built from a rank-overspecified factored representation of the matrix variable.
result Gradient descent iterations converge to the ground-truth matrix under certain conditions and can be stopped efficiently to detect a nearly optimal estimator.
This paper presents a stochastic behavior analysis of a kernel-based stochastic restricted-gradient descent method. The restricted gradient gives a steepest ascent direction within the so-called dictionary subspace. The analysis provides the transient and steady state performance in the mean squared error criterion. It…
This work proposes a new method for variational inference using Wasserstein gradient descent.
problem Optimizing variational parameters to match a true posterior distribution.
method Reinterpreting VI as an optimization problem over a variational parameter space, using Wasserstein gradient descent.
result The proposed Wasserstein gradient descent can be seen as a generalization of existing optimization techniques in VI.
Modern large scale machine learning applications require stochastic optimization algorithms to be implemented on distributed computational architectures. A key bottleneck is the communication overhead for exchanging information such as stochastic gradients among different workers. In this paper, to reduce the communica…
We study two procedures (reverse-mode and forward-mode) for computing the gradient of the validation error with respect to the hyperparameters of any iterative learning algorithm such as stochastic gradient descent. These procedures mirror two methods of computing gradients for recurrent neural networks and have differ…
DoWG optimizer automatically adapts to convex and nonsmooth problems without tuning.
problem Optimizing machine learning models efficiently and adaptively.
method DoWG uses a distance-based weighted version of gradient averaging for optimization.
result DoWG achieves convergence rates similar to optimally tuned gradient descent.
BSVGD improves sampling for multimodal distributions using branching.
problem Sampling from multimodal distributions.
method Random branching in Stein Variational Gradient Descent (SVGD).
result Theoretical convergence guarantee and empirical validation.