ℓ₂-Boosting exhibits benign overfitting with logarithmic excess variance decay Θ(σ²/log(p/n)) under isotropic noise due to ℓ₁ bias, and a subdifferential early stopping rule recovers minimax-optimal ℓ₁ rates.
Canonical reference
Title resolution pending
Canonical reference. 83% of citing Pith papers cite this work as background.
34
Pith papers citing it
Background
83% of classified citations
citation-role summary
background 5
method 1
citation-polarity summary
verdicts
UNVERDICTED 34representative citing papers
The spectral weak-recovery threshold for linearized AMP in the multi-view spiked Wigner model is SNR(λ,B)=1, where SNR is the largest eigenvalue of Diag(√λ)(B⊙B)Diag(√λ), and this coincides with the information-theoretic threshold for a broad class of spike priors.
Proposes pointwise Riemannian Dimension from feature eigenvalues to derive tighter, representation-aware generalization bounds for deep networks in the nonlinear regime.
A projected gradient descent algorithm for noisy inductive matrix completion achieves linear convergence and stable recovery at sample complexity governed by side-information dimension, extending to inexact side-information with optimal error degradation.
TARCO corrects measurement-error-induced correlated contamination in tree-aggregated compositional regression via bias-corrected estimating equations, tree-aware PSD stabilization, and sparse regularization, with finite-sample bounds and sign consistency.
Non-asymptotic decomposition of conditional miscoverage in conformal prediction into score-estimation error, finite-sample calibration error, and intrinsic conditional-mismatch error, with guidance for model selection and extensions to covariate shift and structured data.
The paper establishes the first tilde O(epsilon^{-1}) upper bounds and matching lower bounds for forward-KL-regularized offline contextual bandits under single-policy concentrability in both tabular and general function approximation settings.
Optimistic bilevel optimization with manifold lower-level minimizers is differentiable if the optimistic selection is unique, yielding a pseudoinverse hyper-gradient and a convergent HG-MS algorithm whose rate depends on intrinsic manifold dimension.
ABGD parametrizes piecewise linear functions as difference of max-affine functions and converges linearly to an epsilon-accurate solution with O(d max(sigma/epsilon,1)^2) samples under sub-Gaussian noise, which is minimax optimal up to logs.
A modular reduction from budget-constrained contextual bandits with adversarial contexts to unconstrained bandits via surrogate rewards, yielding improved guarantees and an efficient algorithm based on SquareCB.
A stagewise greedy algorithm for semiparametric contextual dynamic pricing achieves regret T to the max of 1/2 and 3 over (2 beta plus 1) for linear m, with a matching lower bound proving optimality.
The paper proves statistical consistency of contrastive loss to optimal ranking via an AUC criterion and derives generalization bounds O(1/m + 1/sqrt(n)) for supervised and O(1/sqrt(m) + 1/sqrt(n)) for self-supervised CRL that explain benefits of large negative sets.
A Neyman-orthogonal estimator paired with Lasso nuisance estimation achieves root-T asymptotic normality for BLP demand parameters under high-dimensional controls and approximate sparsity.
Bayesian softmax-gated mixture-of-experts models achieve posterior contraction for density estimation and parameter recovery using Voronoi losses, plus two strategies for choosing the number of experts.
RAIC unifies uniform recovery of structured signals from nonlinear observations via PGD, yielding error rates comparable to nonuniform guarantees up to log factors in sparse and 1-bit settings.
A projection-based algorithm for COCO achieves O(log T) regret and O(log T) CCV for strongly convex losses and O(sqrt(T)) for convex losses by leveraging self-contracted curves.
Proposes Factor-Augmented SGD that runs on streaming high-dimensional data and supplies the first convergence analysis explicitly accounting for latent-factor estimation error.
Causal effects are identifiable for categorical unobserved confounders via mixture learning and tensor decomposition, yielding consistent estimators with non-asymptotic guarantees.
Near-linear time algorithm for robust regression under Gaussian covariates achieves O(sqrt(ε κ)) error with Õ(d/ε⁴) samples when ε κ ≲ 1, plus SQ and low-degree lower bounds.
A learning algorithm achieves tight Õ(√T) regret for profit maximization in bilateral trade against smooth adversaries, matching stochastic rates via continuity and algorithmic chaining.
Novel non-asymptotic uniform error bounds are derived for kernel regression under broad classes of non-Gaussian noise distributions that include correlated cases.
dFlowGRPO is a new rate-aware RL method for discrete flow models that outperforms prior GRPO approaches on image generation and matches continuous flow models while supporting broad probability paths.
Safety certification of dynamical systems is reformulated as direct classification via kernel embeddings on trajectories, bypassing recursive DP to avoid error compounding and support non-Markovian dynamics.
MoFI-FLR recovers active covariates and identifies their true functional forms (simple or complex) in high-dimensional functional linear regressions.
citing papers explorer
-
RAFT: Reward rAnked FineTuning for Generative Foundation Model Alignment
RAFT aligns generative models by ranking samples with a reward model and fine-tuning only on the top-ranked outputs, reporting gains on reward scores and automated metrics for LLMs and diffusion models.