Noise-Driven Escape from Metastable Phases explains Grokking in Deep Neural Networks

• URL: http://arxiv.org/abs/2606.17120v1
• Date: Mon, 15 Jun 2026 11:22:16 GMT
• Title: Noise-Driven Escape from Metastable Phases explains Grokking in Deep Neural Networks
• Abstract summary: We show that deep neural networks (DNNs) exhibit first order phase transitions under variations of the L2 regularization strength.<n>We reproduce grokking-like delayed convergence across two orders of magnitude in escape time.<n>Our work suggests that the number of metastable states equals the number of learnable features.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Deep neural networks (DNNs) exhibit first order phase transitions under variations of the L2 regularization strength, with each transition marking the onset of a new learnable feature. Below a critical regularization strength, all features are in principle learnable, but coexisting metastable states, separated by energy barriers, can trap the network and impede convergence. A strength of DNNs is their ability to generalize. But many open questions remain, among them the origin of so called grokking: the abrupt, delayed onset of generalization after prolonged apparent overfitting. We show for linear DNNs that grokking is consistent with hysteresis in first-order L2 phase transitions: using L2 regularization to engineer deliberate trapping, we demonstrate that a model in a low-accuracy metastable state escapes only when SGD noise drives it across an energy barrier, with escape times following Arrhenius scaling. We reproduce grokking-like delayed convergence across two orders of magnitude in escape time by deliberately trapping models in metastable phases. Using sparse sub-sampling we also reproduce the canonical grokking curve where test error eventually approaches the final training error. Our work suggests that the number of metastable states equals the number of learnable features -- one per singular value of the data covariance -- the potential for hysteresis grows naturally with task complexity. We provide evidence that the same mechanism likely operates in general nonlinear DNNs. Our results provide routes toward more efficient learning schemes.

Related papers

• Grokking as a Variance-Limited Phase Transition: Spectral Gating and the Epsilon-Stability Threshold [0.0]
  Generalization requires the textitanisotropic rectification unique to adaptives, which directs noise into the tangent space of the solution.<n>This work analyzes AdamW dynamics on modular arithmetic tasks, revealing a Spectral Gating'' mechanism that regulates the transition from generalization to memorization.
  arXiv  Detail & Related papers  (2026-03-16T16:18:18Z)
• Phase Transitions between Accuracy Regimes in L2 regularized Deep Neural Networks [0.0]
  Increasing the L2 regularization of Deep Neural Networks (DNNs) causes a first-order phase transition into the under-parametrized phase.<n>We predict new transition points as the data complexity is increased and, in accordance with the theory of phase transitions, the existence of effects.
  arXiv  Detail & Related papers  (2025-05-10T11:02:30Z)
• The late-stage training dynamics of (stochastic) subgradient descent on homogeneous neural networks [2.1178416840822027]
  We consider the setting of classification with homogeneous neural networks.<n>We show that normalized SGD iterates converge to the set of critical points of the normalized margin at late-stage training.
  arXiv  Detail & Related papers  (2025-02-08T19:09:16Z)
• Learning with Norm Constrained, Over-parameterized, Two-layer Neural Networks [54.177130905659155]
  Recent studies show that a reproducing kernel Hilbert space (RKHS) is not a suitable space to model functions by neural networks. In this paper, we study a suitable function space for over- parameterized two-layer neural networks with bounded norms.
  arXiv  Detail & Related papers  (2024-04-29T15:04:07Z)
• Benign Overfitting in Deep Neural Networks under Lazy Training [72.28294823115502]
  We show that when the data distribution is well-separated, DNNs can achieve Bayes-optimal test error for classification. Our results indicate that interpolating with smoother functions leads to better generalization.
  arXiv  Detail & Related papers  (2023-05-30T19:37:44Z)
• Theoretical Characterization of the Generalization Performance of Overfitted Meta-Learning [70.52689048213398]
  This paper studies the performance of overfitted meta-learning under a linear regression model with Gaussian features. We find new and interesting properties that do not exist in single-task linear regression. Our analysis suggests that benign overfitting is more significant and easier to observe when the noise and the diversity/fluctuation of the ground truth of each training task are large.
  arXiv  Detail & Related papers  (2023-04-09T20:36:13Z)
• Learning Low Dimensional State Spaces with Overparameterized Recurrent Neural Nets [57.06026574261203]
  We provide theoretical evidence for learning low-dimensional state spaces, which can also model long-term memory. Experiments corroborate our theory, demonstrating extrapolation via learning low-dimensional state spaces with both linear and non-linear RNNs.
  arXiv  Detail & Related papers  (2022-10-25T14:45:15Z)
• Gradient flows and randomised thresholding: sparse inversion and classification [0.0]
  Sparse inversion and classification problems are ubiquitous in modern data science and imaging. In classification, we consider, e.g., the sum of a data fidelity term and a non-smooth Ginzburg--Landau energy. Standard (sub)gradient descent methods have shown to be inefficient when approaching such problems.
  arXiv  Detail & Related papers  (2022-03-22T09:21:14Z)
• Graph-adaptive Rectified Linear Unit for Graph Neural Networks [64.92221119723048]
  Graph Neural Networks (GNNs) have achieved remarkable success by extending traditional convolution to learning on non-Euclidean data. We propose Graph-adaptive Rectified Linear Unit (GReLU) which is a new parametric activation function incorporating the neighborhood information in a novel and efficient way. We conduct comprehensive experiments to show that our plug-and-play GReLU method is efficient and effective given different GNN backbones and various downstream tasks.
  arXiv  Detail & Related papers  (2022-02-13T10:54:59Z)
• On the Double Descent of Random Features Models Trained with SGD [78.0918823643911]
  We study properties of random features (RF) regression in high dimensions optimized by gradient descent (SGD) We derive precise non-asymptotic error bounds of RF regression under both constant and adaptive step-size SGD setting. We observe the double descent phenomenon both theoretically and empirically.
  arXiv  Detail & Related papers  (2021-10-13T17:47:39Z)
• Uncertainty Propagation in Convolutional Neural Networks: Technical Report [13.224071661974593]
  We study the problem of propagation of uncertainty through typical building blocks of a Convolutional Neural Network (CNN) These include layers that perform linear operations, as well as layers that act non-linearly on their input, such as the Rectified Linear Unit (ReLU)
  arXiv  Detail & Related papers  (2021-02-11T15:17:46Z)
• Direction Matters: On the Implicit Bias of Stochastic Gradient Descent with Moderate Learning Rate [105.62979485062756]
  This paper attempts to characterize the particular regularization effect of SGD in the moderate learning rate regime. We show that SGD converges along the large eigenvalue directions of the data matrix, while GD goes after the small eigenvalue directions.
  arXiv  Detail & Related papers  (2020-11-04T21:07:52Z)

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.