Optimistic algorithms achieve logarithmic regret bounds for MDPs without diameter dependence.
problem Achieving logarithmic regret bounds for episodic MDPs without relying on diameter-like quantities.
method Novel 'clipped' regret decomposition applied to optimistic algorithms.
result Smooth interpolation between gap-dependent and minimax rates of convergence.
Two adaptive algorithms improve tracking regret in dynamic expert advice problems.
problem Prediction with expert advice in dynamic environments.
method Developed two adaptive and efficient algorithms using online mirror descent framework.
result Achieved data-dependent tracking regret bounds for both algorithms.
New algorithm balances personalization and statistical validity in MRTs.
problem Optimizing decisions in nonstationary settings with habituation and recovery.
method ROGUE-TS Thompson Sampling with probability clipping.
result Achieves lower regret and maintains high statistical power.
Paper improves worst-case regret bounds for RLSVI in reinforcement learning.
problem Minimizing regret in reinforcement learning with randomized value functions.
method Introduces a clipping variant of Thompson Sampling for RLSVI.
result Achieves a i l d e O ( H 2 S A T ) ilde{\mathrm{O}}(H^2S\sqrt{AT}) i l d e O ( H 2 S A T ) worst-case regret bound. New algorithms reduce regret in online convex optimization with heavy-tailed gradients.
problem Challenges in online convex optimization with heavy-tailed gradients.
method Examined and analyzed old algorithms for online convex optimization in the heavy-tailed setting.
result Established new regret bounds for classical methods without algorithmic modification.
New UCB-type algorithms reduce regret bounds for stochastic bandits with heavy and super heavy noise.
problem Improving regret bounds for stochastic bandits with heavy-tailed noise.
method General convex optimization methods with an inexact oracle, Clipped-SGD-UCB algorithm.
result Achieved an O ( log T K T log T ) O(\log T\sqrt{KT\log T}) O ( log T K T log T ) regret bound for symmetric noise, better than general lower bounds. New learning dynamics achieve fast convergence in games without needing to know utility scales.
problem Fast convergence guarantees in learning games require prior knowledge of utility scales.
method Developed scale-free and scale-invariant learning dynamics using optimistic follow-the-regularized-leader with adaptive learning rates and clipping techniques.
result Achieved fast convergence rates to Nash and correlated equilibria without prior utility scale knowledge.
Paper proposes ClipSMT algorithm for better ATE estimation.
problem Adaptive estimation of Average Treatment Effect (ATE).
method ClipSMT algorithm for improved Neyman regret.
result Achieves exponential improvements in Neyman regret.
This paper improves Q-learning bounds using reference-advantage decomposition.
problem Improving Q-learning bounds in MDPs with positive suboptimality gaps.
method Develops a novel error decomposition framework to prove gap-dependent regret bounds.
result Establishes logarithmic gap-dependent regret bounds for Q-learning.
New algorithm reduces best-in-class regret in contextual bandits.
problem Compete with the best policy in a class without model restrictions.
method Proposes an algorithm that updates policies by minimizing a pessimistic objective, including a clipped inverse-propensity estimate and variance penalty.
result Achieves fast best-in-class regret rates, including polylogarithmic rates in the parametric case.
We introduce a new stochastic smoothing perspective to study adversarial contextual bandit problems. We propose a general algorithm template that represents random perturbation based algorithms and identify several perturbation distributions that lead to strong regret bounds. Using the idea of smoothness, we provide an…
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.
Improved differentially private deep learning with group-wise clipping techniques.
problem Efficiency and privacy trade-offs in deep learning models.
method Group-wise clipping techniques (per-layer and per-device) to reduce compute time and memory overhead.
result Private learning with group-wise clipping achieves similar or better performance than non-private learning with less wall time.
Study on gradient clipping in SGD for high-dimensional problems.
problem Understanding and optimizing gradient clipping in high-dimensional machine learning models.
method Theoretical analysis of streaming SGD with gradient clipping in a least squares problem, focusing on large intrinsic dimensionality.
result Developed a deterministic equation to describe the loss evolution in clipped SGD, showing benefits under certain conditions.
New algorithm improves bandit with graph feedback by decomposing regret.
problem Improving performance in bandit problems with graph feedback.
method Partition-based algorithm framework using regret decomposition.
result Improved and optimal regret bounds on various graph families.
New method improves fairness in DP learning by preventing excessive gradient suppression.
problem Disparate impact on model predictions for minority groups in DP learning.
method Bounded adaptive clipping to prevent excessive gradient suppression.
result Improves worst-class accuracy by over 10 percentage points compared to existing methods.
Paper proposes FedQ-Advantage for federated Q-learning with near-optimal regret and low communication cost.
problem Near-optimal federated Q-learning with low communication cost.
method Reference-advantage decomposition for variance reduction, synchronization between agents and server, policy update.
result Achieves almost optimal regret and near-linear regret speedup compared to single-agent learning.
MOTS improves Thompson sampling to match minimax bounds for bandit problems.
problem Achieving optimal regret bounds for Thompson sampling in multi-armed bandit problems.
method Proposes MOTS, a variant of Thompson sampling that clips arm sampling.
result Proves MOTS achieves minimax optimal regret bounds O ( K T ) O(\sqrt{KT}) O ( K T ) . Private adaptive methods improve on traditional SGD for convex optimization.
problem Differential privacy constraints in gradient optimization.
method Differentially private variants of SGD and AdaGrad with adaptive stepsizes and non-isotropic clipping.
result Private AdaGrad outperforms private SGD in high-dimensional problems.
This paper analyzes convergence of DP-SGD with adaptive quantile clipping.
problem Empirical success of adaptive clipping methods lacks theoretical understanding.
method Comprehensive convergence analysis of SGD with quantile clipping (QC-SGD).
result Establishes theoretical guarantees for DP-QC-SGD, revealing relationships between quantile selection, step size, and convergence.
A new method improves training speed and robustness for deep learning models.
problem Training deep learning models with heavy-tailed gradient noise.
method Per-sample gradient clipping combined with SGD.
result PS-Clip-SGD achieves optimal convergence rates and high-probability guarantees.
CLIP models robustness to spurious features is re-evaluated using a new dataset.
problem Existing robustness tests of CLIP models may not fully reflect their performance on spurious features.
method Crafted a new dataset (CounterAnimal) to reveal CLIP models' reliance on realistic spurious features.
result CLIP models are robust to spurious features learned from their training data, not ImageNet.
CLIP learns joint image-text representations for zero-shot learning.
problem Understanding and improving zero-shot transfer performance in CLIP.
method Formal study of transferrable representation learning and analysis of zero-shot transfer performance.
result Proposes a new CLIP-type approach that outperforms existing methods.
Adaptive clipping improves privacy in federated learning.
problem Inadequate fixed clipping norms in existing DP Federated Averaging methods.
method Adaptive clipping to a quantile of the update norm distribution, estimated online with differential privacy.
result Adaptive clipping to the median update norm outperforms fixed clipping in federated learning tasks.
Paper introduces Latent-CLIP for efficient text-image comparison in latent space.
problem Efficiently compare text and images in latent space without costly decoding.
method Trains CLIP model in latent space, uses Latent-CLIP rewards for noise optimization, and guides generation away from harmful content.
result Latent-CLIP matches CLIP performance on text-image classification and harmful content detection.
Gradient clipping helps private SGD converge despite potential bias.
problem Gradient clipping in private SGD can bias convergence.
method Theoretical analysis and empirical evaluation of gradient clipping effects.
result Gradient clipping can prevent convergence to stationary points and introduces bias.
This paper improves analysis of clipping algorithms for non-convex optimization.
problem Improving convergence analysis of clipping algorithms for non-convex optimization.
method Develops a general framework to study clipping algorithms, including momentum methods, and provides convergence analysis in deterministic and stochastic settings.
result Demonstrates that clipping methods can maintain efficiency even in highly non-smooth regions.
Enhances CLIP's similarity computation using PMI's linear structure.
problem CLIP's similarity computation misses the optimal linear structure of PMI.
method KME-CLIP, utilizing inner product in a reproducing kernel Hilbert space.
result KME-CLIP approximates PMI with arbitrary accuracy and outperforms CLIP.
AutoClip automatically adjusts gradient clipping for better audio separation.
problem Improving generalization in audio source separation networks.
method Adaptive gradient clipping based on historical gradient norms.
result Improves generalization performance in audio source separation networks.
Dp-CLIP preserves privacy in multimodal AI training.
problem Privacy concerns in multimodal AI, especially in vision-language tasks.
method Differentially private adaptation of CLIP model.
result Dp-CLIP retains accuracy while ensuring privacy.
New algorithm reduces reinforcement learning regret to sqrt(T) without strong dynamics assumptions.
problem Infinite-horizon average-reward reinforcement learning with linear MDPs.
method Approximate by discounted-reward MDPs and apply optimistic value iteration.
result Achieves O(sqrt(T)) regret with polynomial complexity.
This paper tackles minimizing clipped convex functions with heuristics and mixed-integer convex programming.
problem Minimizing a sum of clipped convex functions.
method Heuristics and mixed-integer convex programming.
result Heuristics can find good solutions, and the perspective transformation yields tractable lower bounds.
This paper improves convergence guarantees for gradient clipping in deep learning.
problem Improving convergence guarantees for gradient clipping in deep learning models.
method Analyzes and provides precise convergence guarantees for arbitrary clipping thresholds.
result Shows tight convergence guarantees for clipped stochastic gradient descent.
Unified CLIP space manipulations improve GAN adaptation with a single target image.
problem Overfitting or underfitting in fine-tuning a pre-trained generator with a single target image.
method Two-step training strategy: latent optimization in CLIP space followed by generator fine-tuning with CLIP space consistency loss.
result Our model generates diverse outputs with the target texture and outperforms baseline models.
Improved online convex optimization with delayed feedback using curvature.
problem Online convex optimization with curved losses and delayed feedback.
method Variant of follow-the-regularized-leader and Online Newton Step algorithm with adaptive learning rate.
result Regret bounds of order min { σ max ln T , d t o t } \min\{σ_{\max}\ln T, \sqrt{d_{\mathrm{tot}}}\} min { σ m a x ln T , d tot } for exp-concave losses. Clip21 improves convergence of gradient-clipped methods in DP settings.
problem Gradient clipping introduces bias in distributed training, causing convergence issues.
method Clip21 designs an error feedback mechanism to mitigate bias in gradient-clipped methods.
result Clip21 converges at the same rate as distributed gradient descent, improving from previous $\mathcal{O}\left(\frac{1}{\sqrt{K}}
ight)$ to $\mathcal{O}\left(\frac{1}{K}
ight)$ .
New method for estimating covariance with robustness to outliers.
problem Estimating covariance from noisy data with outliers.
method Cross-fitted clipped covariance estimator with computable Bernstein certificates.
result The method balances certified stochastic error and robust hold-out proxy for clipping bias.
This paper begins with a study on the dual representations of risk and regret measures and their impact on modeling multistage decision making under uncertainty. A relationship between risk envelopes and regret envelopes is established by using the Lagrangian duality theory. Such a relationship opens a door to a decomp…
Paper proves convergence for private FL on non-Lipschitz convex objectives using normalization instead of clipping.
problem Lack of convergence results for differentially private federated learning with non-Lipschitz objectives.
method Developed a convergence result for private FL on smooth convex objectives without assuming Lipschitzness, using normalization instead of clipping.
result Normalization-based private FL algorithm converges better than clipping-based counterpart on smooth convex functions.
INF-clip optimizes heavy-tailed MAB problems with improved performance.
problem Optimizing multi-armed bandit problems with heavy-tailed rewards.
method INF-clip algorithm for adversarial and stochastic heavy-tailed MAB settings.
result INF-clip is optimal for linear and non-linear heavy-tailed stochastic MAB problems.
Paper explores whether gradient normalization can replace clipping for SGD in heavy-tailed noise.
problem Ensuring convergence of SGD in heavy-tailed noise.
method Revisits gradient clipping and normalization, proving their sufficiency and effectiveness.
result Gradient normalization alone is sufficient for nonconvex SGD convergence under smoothness assumptions.
Many continuous control tasks have bounded action spaces. When policy gradient methods are applied to such tasks, out-of-bound actions need to be clipped before execution, while policies are usually optimized as if the actions are not clipped. We propose a policy gradient estimator that exploits the knowledge of action…
The paper introduces new measures to quantify variability in decision tree models due to observational multiplicity.
problem The variability in decision tree models due to observational multiplicity.
method Introduces leaf regret and structural regret to decompose observational multiplicity.
result Structural regret is the primary driver of observational multiplicity, accounting for over 15 times the variability of leaf regret in some datasets.
TROLL improves RL for LLMs by replacing clipping with a trust region projection.
problem Clipping in RL for LLMs causes instability and suboptimal performance.
method TROLL uses a discrete differentiable trust region projection to replace clipping, balancing computational cost and effectiveness.
result TROLL consistently outperforms PPO-like clipping in training speed, stability, and final success rates.
New analysis shows how to balance privacy and accuracy in deep learning.
problem Balancing privacy and accuracy in deep learning models.
method Continuous time analysis through neural tangent kernel (NTK) for arbitrary architectures.
result Large clipping norm improves calibration without sacrificing accuracy.
New algorithm achieves asymptotically optimal regret without horizon dependence.
problem Horizon-free regret minimization for reinforcement learning.
method Proposes a new algorithm and proves a regret upper bound.
result Regret upper bound of \(\tilde O(\sqrt{SAK} + S^8A^3)\) with failure probability \(\delta\).
Robust CLIP improves vision models' resistance to attacks.
problem Vulnerability of vision models to adversarial attacks.
method Unsupervised adversarial fine-tuning of CLIP.
result Robust CLIP prevents stealth attacks on vision models.
Theory for RLHF generalization under reward shift and clipped KL.
problem Theoretical understanding of RLHF generalization, especially with reward shift and clipped KL.
method Developed generalization theory for RLHF, accounting for reward shift and clipped KL.
result Presented generalization bounds for RLHF, suggesting generalization error from sampling, reward shift, and KL clipping.