SoftSignSGD(S3): An Enhanced Optimizer for Practical DNN Training and Loss Spikes Minimization Beyond Adam

• URL: http://arxiv.org/abs/2507.06464v1
• Date: Wed, 09 Jul 2025 00:47:37 GMT
• Title: SoftSignSGD(S3): An Enhanced Optimizer for Practical DNN Training and Loss Spikes Minimization Beyond Adam
• Abstract summary: Adam has proven successful in training deep neural networks, but the mechanisms underlying its successes and limitations remain underexplored.<n>In this study, we demonstrate that the effectiveness of Adam stems largely from its similarity to SignSGD in robustly fluctuating fluctuations.<n>We propose $emphThird, S3 incorporates an equivalent Nesterov's accelerated gradient(NAG) module, accelerating convergence without memory overhead.
• Score: 28.46753383125989
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Adam has proven remarkable successful in training deep neural networks, but the mechanisms underlying its empirical successes and limitations remain underexplored. In this study, we demonstrate that the effectiveness of Adam stems largely from its similarity to SignSGD in robustly handling large gradient fluctuations, yet it is also vulnerable to destabilizing loss spikes due to its uncontrolled update scaling. To enhance the advantage of Adam and mitigate its limitation, we propose SignSoftSGD (S3), a novel optimizer with three key innovations. \emph{First}, S3 generalizes the sign-like update by employing a flexible $p$-th order momentum ($p \geq 1$) in the denominator, departing from the conventional second-order momentum (variance) preconditioning. This design enables enhanced performance while achieving stable training even with aggressive learning rates. \emph{Second}, S3 minimizes the occurrences of loss spikes through unified exponential moving average coefficients for numerator and denominator momenta, which inherently bound updates to $[-1, 1]$ and simplify hyperparameter tuning. \emph{Third}, S3 incorporates an equivalent Nesterov's accelerated gradient(NAG) module, accelerating convergence without memory overhead. Theoretically, we prove that S3 achieves the optimal convergence rate of $O\left(\frac{1}{T^{\sfrac{1}{4}}}\right)$ for general nonconvex stochastic optimization under weak assumptions. Extensive experiments across a range of vision and language tasks show that \textsf{\small S3} not only converges more rapidly and improves performance but also rarely experiences loss spikes, even with a \textbf{$\bm{10 \times}$} larger learning rate. In fact, S3 delivers performance comparable to or better than AdamW with \textbf{$2 \times$} the training steps, establishing its efficacy in both efficiency and final task performance.

Related papers

• Singularity-aware Optimization via Randomized Geometric Probing: Towards Stable Non-smooth Optimization [8.412194372218895]
  We introduce Singularity-aware Adam (S-Adam), a novel that stabilizes training by dynamically modulating step sizes based on local geometric instability.<n>S-Adam incorporates an adaptive damping mechanism that decelerates updates in high-instability regions while preserving fast convergence in smooth basins.<n> Empirical evaluations on optimal Quantization-Aware Training (QAT) and high-noise small-batch learning demonstrate that S-Adam consistently outperforms AdamW and Prox-SGD.
  arXiv  Detail & Related papers  (2026-05-28T08:00:40Z)
• Beyond Heuristics: Learnable Density Control for 3D Gaussian Splatting [78.96382179408054]
  We introduce bfLeGS, a framework that reformulates density control as a parameterized policy network optimized via Reinforcement Learning (RL)<n>To maintain computational tractability, we derive a closed-form solution that reduces the complexity of reward calculation from $O(N2)$ to $O(N)$.<n>Experiments on the Mip-NeRF 360, Tanks & Temples, and Deep Blending datasets demonstrate that bfLeGS significantly outperforms state-of-the-art methods.
  arXiv  Detail & Related papers  (2026-05-01T05:04:49Z)
• EquiformerV3: Scaling Efficient, Expressive, and General SE(3)-Equivariant Graph Attention Transformers [70.68282700439444]
  We introduce EquiformerV3, the third generation of the $SE(3)$-equivariant graph attention Transformer.<n>We optimize the software implementation, achieving $1.75times$ speedup.<n>We propose SwiGLU-$S2$ activations to incorporate many-body interactions for better theoretical expressivity.
  arXiv  Detail & Related papers  (2026-04-10T09:12:16Z)
• Fundamental Limits of Neural Network Sparsification: Evidence from Catastrophic Interpretability Collapse [7.167095059974211]
  This work investigates feature survival under severe capacity constraints in Variational Autoencoder--Sparse Autoencoder architectures.<n>We introduce an adaptive sparsity scheduling framework that progressively reduces active neurons from 500 to 50 over 50 training epochs.<n>Our key finding reveals a pervasive paradox: while global representation quality remains stable, local feature interpretability collapses systematically.
  arXiv  Detail & Related papers  (2026-03-18T00:16:38Z)
• Causal Direction from Convergence Time: Faster Training in the True Causal Direction [0.0]
  We introduce Causal Computational Asymmetry (CCA), a principle for causal direction identification based on optimization dynamics.<n>CCA operates in optimization-time space, distinguishing it from methods such as RESIT, IGCI, and SkewScore.<n>We further embed CCA into a broader framework termed Causal Compression Learning (CCL), which integrates graph structure learning, causal information compression, and policy optimization.
  arXiv  Detail & Related papers  (2026-02-24T21:34:57Z)
• Time Is All It Takes: Spike-Retiming Attacks on Event-Driven Spiking Neural Networks [87.16809558673403]
  Spiking neural networks (SNNs) compute with discrete spikes and exploit temporal structure.<n>We study a timing-only adversary that retimes existing spikes while preserving spike counts and amplitudes in event-driven SNNs.
  arXiv  Detail & Related papers  (2026-02-03T09:06:53Z)
• AdamHD: Decoupled Huber Decay Regularization for Language Model Pre-Training [0.2578242050187029]
  AdamHuberDecay is a drop-in replacement for AdamW that substitutes the $ell$ penalty with a decoupled smooth Huber regularizer.<n>Experiments on GPT-2 and GPT-3 pre-training demonstrate that AdamHuberDecay converges 10-15% faster in wall-clock time.
  arXiv  Detail & Related papers  (2025-11-18T18:08:20Z)
• Optimization and generalization analysis for two-layer physics-informed neural networks without over-parametrization [0.6215404942415159]
  This work focuses on the behavior of gradient descent (SGD) in solving least-squares regression with physics-informed neural networks (PINNs)<n>We show that if the network width exceeds a threshold that depends only on $epsilon$ and the problem, then the training loss and expected loss will decrease below $O(epsilon)$.
  arXiv  Detail & Related papers  (2025-07-22T09:24:22Z)
• Non-stationary Online Learning for Curved Losses: Improved Dynamic Regret via Mixability [65.99855403424979]
  We show that dynamic regret can be substantially improved by leveraging the concept of mixability.<n>We demonstrate that an exponential-weight method with fixed-share updates achieves an $mathcalO(d T2/3 P_T2/3 log T)$ dynamic regret for mixable losses.
  arXiv  Detail & Related papers  (2025-06-12T12:00:08Z)
• VAMO: Efficient Large-Scale Nonconvex Optimization via Adaptive Zeroth Order Variance Reduction [3.130722489512822]
  VAMO combines FO mini-batch gradients with ZO finite-difference probes under an ZOG-style framework.<n>VAMO outperforms established FO and ZO methods, offering a faster, more flexible option for improved efficiency.
  arXiv  Detail & Related papers  (2025-05-20T05:31:15Z)
• Pushing the Limits of Low-Bit Optimizers: A Focus on EMA Dynamics [64.62231094774211]
  Statefuls (e.g., Adam) maintain auxiliary information even 2x the model size in order to achieve optimal convergence.<n>SOLO enables Adam-styles to maintain quantized states with precision as low as 3 bits, or even 2 bits.<n>SOLO can thus be seamlessly applied to Adam-styles, leading to substantial memory savings with minimal accuracy loss.
  arXiv  Detail & Related papers  (2025-05-01T06:47:45Z)
• Second-order Optimization of Gaussian Splats with Importance Sampling [51.95046424364725]
  3D Gaussian Splatting (3DGS) is widely used for novel view rendering due to its high quality and fast inference time.<n>We propose a novel second-order optimization strategy based on Levenberg-Marquardt (LM) and Conjugate Gradient (CG)<n>Our method achieves a $3times$ speedup over standard LM and outperforms Adam by $6times$ when the Gaussian count is low.
  arXiv  Detail & Related papers  (2025-04-17T12:52:08Z)
• TeZO: Empowering the Low-Rankness on the Temporal Dimension in the Zeroth-Order Optimization for Fine-tuning LLMs [58.19080159470868]
  We propose a novel low-rank ZO estimator, TeZO, which captures the low-rankness across both the model and temporal dimension.<n>Specifically, we represent ZO perturbations along the temporal dimension as a 3D tensor and employ Canonical Polyadic Decomposition (CPD) to extract each low-rank 2D matrix.
  arXiv  Detail & Related papers  (2025-01-31T11:34:03Z)
• Taming 3DGS: High-Quality Radiance Fields with Limited Resources [50.92437599516609]
  3D Gaussian Splatting (3DGS) has transformed novel-view synthesis with its fast, interpretable, and high-fidelity rendering. We tackle the challenges of training and rendering 3DGS models on a budget. We derive faster, numerically equivalent solutions for gradient computation and attribute updates.
  arXiv  Detail & Related papers  (2024-06-21T20:44:23Z)
• Variance-reduced Clipping for Non-convex Optimization [24.765794811146144]
  Gradient clipping is a technique used in deep learning applications such as large-scale language modeling. Recent experimental training have a fairly special behavior in that it mitigates order complexity.
  arXiv  Detail & Related papers  (2023-03-02T00:57:38Z)
• Flatter, faster: scaling momentum for optimal speedup of SGD [0.0]
  We study training dynamics arising from interplay between gradient descent (SGD) and label noise and momentum in the training of neural networks. We find that scaling the momentum hyper parameter $1-NISTbeta$ with the learning rate to the power of $2/3$ maximally accelerates training, without sacrificing generalization.
  arXiv  Detail & Related papers  (2022-10-28T20:41:48Z)
• Confident Adaptive Language Modeling [95.45272377648773]
  CALM is a framework for dynamically allocating different amounts of compute per input and generation timestep. We demonstrate the efficacy of our framework in reducing compute -- potential speedup of up to $times 3$ -- while provably maintaining high performance.
  arXiv  Detail & Related papers  (2022-07-14T17:00:19Z)
• Towards Noise-adaptive, Problem-adaptive Stochastic Gradient Descent [7.176107039687231]
  We design step-size schemes that make gradient descent (SGD) adaptive to (i) the noise. We prove that $T$ iterations of SGD with Nesterov iterations can be near optimal. Compared to other step-size schemes, we demonstrate the effectiveness of a novel novel exponential step-size scheme.
  arXiv  Detail & Related papers  (2021-10-21T19:22:14Z)
• Momentum Improves Normalized SGD [51.27183254738711]
  We show that adding momentum provably removes the need for large batch sizes on objectives. We show that our method is effective when employed on popular large scale tasks such as ResNet-50 and BERT pretraining.
  arXiv  Detail & Related papers  (2020-02-09T07:00:54Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.