arXiv research

Improved differentially private deep learning with group-wise clipping techniques.

This paper analyzes convergence of DP-SGD with adaptive quantile clipping.

Improved optimization technique reduces training complexity for non-convex problems.

Gradient clipping helps private SGD converge despite potential bias.

This paper improves analysis of clipping algorithms for non-convex optimization.

In importance sampling (IS)-based reinforcement learning algorithms such as Proximal Policy Optimization (PPO), IS weights are typically clipped to avoid large variance in learning. However, policy update from clipped statistics induces large bias in tasks with high action dimensions, and bias from clipping makes it di…

A new algorithm reduces communication in deep learning training.

U-Clip clips gradients to reduce bias without discarding them.

Deep Neural Networks (DNNs) are widely being adopted for safety-critical applications, e.g., healthcare and autonomous driving. Inherently, they are considered to be highly error-tolerant. However, recent studies have shown that hardware faults that impact the parameters of a DNN (e.g., weights) can have drastic impact…

This paper analyzes bias-variance trade-off for clipped SFOMs, improving complexity guarantees for heavy-tailed noise.

Study on gradient clipping in SGD for high-dimensional problems.

PPO's gradients are heavy-tailed, affecting learning; a robust estimator improves performance.

New method improves fairness in DP learning by preventing excessive gradient suppression.

New methods speed up training of differentially private deep learning models.

A new method improves training speed and robustness for deep learning models.

CLIP models robustness to spurious features is re-evaluated using a new dataset.

CLIP learns joint image-text representations for zero-shot learning.

Paper introduces Latent-CLIP for efficient text-image comparison in latent space.

Enhances CLIP's similarity computation using PMI's linear structure.

AutoClip automatically adjusts gradient clipping for better audio separation.

Dp-CLIP preserves privacy in multimodal AI training.

This paper improves convergence guarantees for gradient clipping in deep learning.

Unified CLIP space manipulations improve GAN adaptation with a single target image.

Existing approaches for training neural networks with user-level differential privacy (e.g., DP Federated Averaging) in federated learning (FL) settings involve bounding the contribution of each user's model update by clipping it to some constant value. However there is no good a priori setting of the clipping norm acr…

New algorithm maintains privacy while improving model performance in selective release.

Clip21 improves convergence of gradient-clipped methods in DP settings.

New method for estimating covariance with robustness to outliers.

This paper is concerned with offline reinforcement learning (RL), which learns using pre-collected data without further exploration. Effective offline RL would be able to accommodate distribution shift and limited data coverage. However, prior algorithms or analyses either suffer from suboptimal sample complexities or …

We establish the first benchmark for federated learning with differential privacy in ASR, achieving competitive performance.

Paper proves convergence for private FL on non-Lipschitz convex objectives using normalization instead of clipping.

INF-clip optimizes heavy-tailed MAB problems with improved performance.

Paper explores whether gradient normalization can replace clipping for SGD in heavy-tailed noise.

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…

New algorithm for differentially private distributed optimization of smooth, non-convex problems.

TROLL improves RL for LLMs by replacing clipping with a trust region projection.

New analysis shows how to balance privacy and accuracy in deep learning.

Recently, the study of heavy-tailed noises in first-order nonconvex stochastic optimization has gotten a lot of attention since it was recognized as a more realistic condition as suggested by many empirical observations. Specifically, the stochastic noise (the difference between the stochastic and true gradient) is con…

We consider the problem of minimizing a sum of clipped convex functions; applications include clipped empirical risk minimization and clipped control. While the problem of minimizing the sum of clipped convex functions is NP-hard, we present some heuristics for approximately solving instances of these problems. These h…

Robust CLIP improves vision models' resistance to attacks.

Theory for RLHF generalization under reward shift and clipped KL.

Enhanced metrics improve generative model evaluation reliability.

Improved Clipped-SGD achieves near-optimal heavy-tailed statistical estimation in streaming settings.

New optimization method combines gradient clipping and non-Euclidean smoothness.

We consider the problem of recovering a low-rank matrix from its clipped observations. Clipping is conceivable in many scientific areas that obstructs statistical analyses. On the other hand, matrix completion (MC) methods can recover a low-rank matrix from various information deficits by using the principle of low-ran…

New method improves convergence of SGD for heavy-tailed noise.

New learning dynamics achieve fast convergence in games without needing to know utility scales.

Paper compares RL models for finance, finding Reward Clipping best.

Parametric t-SNE improves generalization for streaming data.