ALRC clips learning rates to stabilize neural network training.
problem Stability issues in training with small batch sizes and high learning rates.
method Adaptive learning rate clipping (ALRC) to limit losses.
result ALRC decreases errors in unstable training conditions.
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.
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…
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 $\mathcal{O}\left(\frac{1}{K}
ight)$ .
Improved optimization technique reduces training complexity for non-convex problems.
problem Training non-convex optimization problems with exploding gradients.
method Employed variance reduction technique (SPIDER) with carefully designed learning rate.
result Improved stochastic gradient complexity to O ( ε − 3 ) O(ε^{-3}) O ( ε − 3 ) for ε ε ε -stationary solutions. 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 optimization method combines gradient clipping and non-Euclidean smoothness.
problem Improving optimization in non-Euclidean spaces for machine learning.
method Hybrid of steepest descent and conditional gradient, incorporating weight decay.
result Achieves optimal convergence rate and demonstrates effectiveness in deep learning.
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.
Improved analysis for clipped gradient methods in nonsmooth convex optimization under heavy-tailed noise.
problem Optimization under heavy-tailed noise in nonsmooth convex problems.
method Refined analysis of Clipped Stochastic Gradient Descent (Clipped SGD) with new rates and improved utilization of Freedman's inequality.
result New rates O ( σ l d m e f f − 1 / 2 p ln 1 − 1 / p ( 1 / δ ) T 1 / p − 1 ) {\cal O}(σ_{\frak l}d_{
m eff}^{-1/2{\frak p}}\ln^{1-1/{\frak p}}(1/δ)T^{1/{\frak p}-1}) O ( σ l d m e f f − 1/2 p ln 1 − 1/ p ( 1/ δ ) T 1/ p − 1 ) and O ( σ l 2 d m e f f − 1 / p ln 2 − 2 / p ( 1 / δ ) T 2 / p − 2 ) {\cal O}(σ_{\frak l}^2d_{
m eff}^{-1/{\frak p}}\ln^{2-2/{\frak p}}(1/δ)T^{2/{\frak p}-2}) O ( σ l 2 d m e f f − 1/ p ln 2 − 2/ p ( 1/ δ ) T 2/ p − 2 ) for nonsmooth convex and strongly convex problems, respectively. Improved Clipped-SGD achieves near-optimal heavy-tailed statistical estimation in streaming settings.
problem High-dimensional heavy-tailed statistical estimation in streaming with memory constraints.
method Stochastic convex optimization with Clipped-SGD, proving near-optimal sub-Gaussian statistical rates.
result Clipped-SGD achieves an error of T r ( Σ ) + T r ( Σ ) ∥ Σ ∥ 2 log ( log ( T ) δ ) T \sqrt{\frac{\mathsf{Tr}(Σ)+\sqrt{\mathsf{Tr}(Σ)\|Σ\|_2}\log(\frac{\log(T)}δ)}{T}} T Tr ( Σ ) + Tr ( Σ ) ∥Σ ∥ 2 l o g ( δ l o g ( T ) ) with probability 1 − δ 1-δ 1 − δ . New method improves optimization and DP in FL.
problem Combining strong DP and optimization in FL.
method Combining clipping, momentum, and error feedback.
result Optimal convergence rate and near optimal DP guarantees.
Optimal DP-GD achieves high-dimensional linear regression risk.
problem Differentially private linear regression in high-dimensional settings.
method One-pass DP gradient descent (DP-GD) with gradient clipping and decaying learning rate.
result Achieves optimal non-asymptotic risk of O ( γ + γ 2 / ρ 2 ) O(γ+ γ^2 / ρ^2) O ( γ + γ 2 / ρ 2 ) for well-conditioned data. 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.
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.
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.
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.
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.
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.
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.
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…
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…
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.
New algorithm for differentially private distributed optimization of smooth, non-convex problems.
problem No differentially private distributed method for smooth, non-convex optimization problems.
method Smoothed normalization integrated with an error-feedback mechanism.
result Achieves superior convergence rate and first differentially private distributed optimization algorithm with provable convergence guarantees.
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.
Paper improves Fisher information estimation methods.
problem Estimating Fisher information for location parameters.
method Revisits and improves Bhattacharya estimator, introduces clipped estimator.
result Clipped estimator shows superior convergence rates in Gaussian noise.
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.
New theory explains why normalization is preferred in SGD under heavy-tailed noise.
problem Understanding why normalization is preferred in stochastic gradient descent (SGD) under heavy-tailed noise.
method Developed a worst-case complexity theory for stochastically preconditioned SGD and its variants.
result Normalization guarantees convergence at optimal rates, while clipping may fail in the worst case.
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.
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.
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.
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.
Paper proposes DP-SGD and DP-NSGD for differentially private non-convex optimization.
problem Mitigating privacy risks in large model learning.
method Clip or normalize per-sample gradients and add noise for differential privacy.
result Achieved convergence rate of gradient norm for non-convex optimization.
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.
Paper compares RL models for finance, finding Reward Clipping best.
problem Optimizing asset allocation in finance.
method Actor-only, actor-critic, and PPO models compared; Reward Clipping introduced.
result Reward Clipping model outperforms others in bull and bear markets.
Spectral algorithms improve under covariate shift with novel weighted techniques.
problem Improving spectral algorithms' performance under covariate shift.
method Analysis of spectral algorithms in non-parametric regression over RKHS, proposing a weighted spectral algorithm with clipped weights.
result Normalized weighted spectral algorithm achieves optimal capacity-independent convergence rates, and clipped weights can approach optimal capacity-dependent rates.
Mitigates faults in DNNs by clipping activation values, improving their resilience.
problem Fault tolerance of DNNs in safety-critical applications.
method Clipping activation functions to reduce impact of faulty weights.
result Significant improvement in classification accuracy (68.92%) for fault mitigation.
Study finds CLIP's caption-based learning outperforms image-only methods under certain conditions.
problem Comparing CLIP's performance with traditional image-only methods in learning transferable representations.
method Controlled comparison of CLIP and image-only methods using a dataset with descriptive captions and specific criteria.
result CLIP's caption-based learning outperforms image-only methods when certain conditions are met, but can be detrimental in others.
Very recently proximal policy optimization (PPO) algorithms have been proposed as first-order optimization methods for effective reinforcement learning. While PPO is inspired by the same learning theory that justifies trust region policy optimization (TRPO), PPO substantially simplifies algorithm design and improves da…
We establish the first benchmark for federated learning with differential privacy in ASR, achieving competitive performance.
problem Challenges in training large transformer models for ASR in federated learning with differential privacy.
method Per-layer clipping and layer-wise gradient normalization to mitigate gradient heterogeneity.
result FL with DP is viable in ASR with strong privacy guarantees, achieving competitive performance.
Most learning algorithms are not invariant to the scale of the function that is being approximated. We propose to adaptively normalize the targets used in learning. This is useful in value-based reinforcement learning, where the magnitude of appropriate value approximations can change over time when we update the polic…
PPOS improves PPO by smoothing the surrogate objective function.
problem Performance instability and optimization inefficiency in PPO.
method Use of a functional clipping method instead of a flat clipping method.
result PPOS conducts more accurate updates and outperforms other PPO variants.
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.
New convergence bounds for online learning with heavy-tailed noise.
problem Learning on streaming data with heavy-tailed noise.
method Nonlinear stochastic gradient descent (SGD) for non-convex and strongly convex costs.
result Strong convergence rates for various nonlinearities and noise distributions.
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.
Paper proposes an active learning method for surgical workflow recognition using long-range temporal dependency.
problem Challenges in automatic surgical workflow recognition due to lack of large-scale labelled datasets.
method NL-RCNet with non-local block for capturing long-range temporal dependency and intra-clip dependency score for selection.
result Our approach outperforms state-of-the-art methods by selecting only 50% of samples for training.
AdaCliP reduces noise in private SGD training.
problem Privacy preserving machine learning over user data.
method Adaptive clipping of gradients to reduce noise in private SGD.
result AdaCliP adds less noise and improves model accuracy.
This paper analyzes bias-variance trade-off for clipped SFOMs, improving complexity guarantees for heavy-tailed noise.
problem Improving complexity guarantees for stochastic optimization methods with heavy-tailed noise.
method Novel analysis of bias-variance trade-off in gradient clipping for clipped SFOMs.
result Improved complexity guarantees for clipped SFOMs across various tail indices, including infinite mean noise.
New method improves convergence of SGD for heavy-tailed noise.
problem Heavy-tailed noise in machine learning applications.
method Normalized SGD (NSGD) to handle heavy-tailed noise.
result Improved sample complexity and high-probability convergence.