Asymptotic analysis of shallow and deep forgetting in replay with Neural Collapse

• URL: http://arxiv.org/abs/2512.07400v1
• Date: Mon, 08 Dec 2025 10:35:57 GMT
• Title: Asymptotic analysis of shallow and deep forgetting in replay with Neural Collapse
• Abstract summary: A persistent paradox in continual learning (CL) is that neural networks often retain linearly separable representations of past tasks even when their output predictions fail.<n>We reveal a critical asymmetry in Experience Replay: while minimal buffers successfully anchor feature geometry, mitigating shallow forgetting typically requires substantially larger buffers.
• Score: 32.34050220649143
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: A persistent paradox in continual learning (CL) is that neural networks often retain linearly separable representations of past tasks even when their output predictions fail. We formalize this distinction as the gap between deep feature-space and shallow classifier-level forgetting. We reveal a critical asymmetry in Experience Replay: while minimal buffers successfully anchor feature geometry and prevent deep forgetting, mitigating shallow forgetting typically requires substantially larger buffer capacities. To explain this, we extend the Neural Collapse framework to the sequential setting. We characterize deep forgetting as a geometric drift toward out-of-distribution subspaces and prove that any non-zero replay fraction asymptotically guarantees the retention of linear separability. Conversely, we identify that the "strong collapse" induced by small buffers leads to rank-deficient covariances and inflated class means, effectively blinding the classifier to true population boundaries. By unifying CL with out-of-distribution detection, our work challenges the prevailing reliance on large buffers, suggesting that explicitly correcting these statistical artifacts could unlock robust performance with minimal replay.

Related papers

• Across the Loss Landscape with Progressive Growth [51.366966420881646]
  We show how grow-and-optimize strategies to train toward flatter regions by viewing growth as progressive relaxation.<n>We empirically validate these predictions in controlled toy basins and in a realistic ResNet/CI-100 setting.<n>Although progressive subspace growth reliably produces flatter solutions, curvature reductions do not universally translate into improved test performance, highlighting subtleties in the flatness-generalization connection.
  arXiv  Detail & Related papers  (2026-08-25T13:50:48Z)
• Drift and Dependence: Layer-wise Information-Theoretic Bounds for Replay-Based Continual Learning [41.371439825328544]
  Continual learning must absorb new tasks without erasing old ones, and replay---mixing a small buffer of past examples into current training---is among the most effective remedies for catastrophic forgetting.<n>We develop a layer-wise information-theoretic framework that separates these effects at every depth.<n>Controlled and benchmark experiments confirm the predicted memory scaling, the interior funnel, and the alignment signal's link to forgetting.
  arXiv  Detail & Related papers  (2026-08-12T06:01:49Z)
• Where to Refine, When to Stop: Rethinking Redundancy via Latent Discrepancy for Efficient Visual Autoregressive Generation [50.139984798361375]
  We propose a training-free framework that removes redundancy via latent discrepancy.<n> LD-Pruning substantially reduces inference while maintaining high generation quality, achieving up to 2.35x speedup on Infinity-8B.
  arXiv  Detail & Related papers  (2026-05-29T19:34:39Z)
• Clipping Bottleneck: Stabilizing RLVR via Stochastic Recovery of Near-Boundary Signals [83.0127582612634]
  Near-boundary Rescue (NSR) is a minimal, plug-and-play modification that retains out-of-bound tokens to recover lost signals.<n>NSR substantially improves training stability and delivers consistent gains over strong baselines such as DAPO and GSPO.
  arXiv  Detail & Related papers  (2026-05-21T16:45:31Z)
• Understanding Scaling Laws in Deep Neural Networks via Feature Learning Dynamics [9.885471525709113]
  We show that scaling laws describe what success looks like but not when and why scaling succeeds or fails.<n>A central obstacle is the lack of a rigorous understanding of feature learning at large depth.
  arXiv  Detail & Related papers  (2025-12-24T09:39:04Z)
• SPEAR++: Scaling Gradient Inversion via Sparsely-Used Dictionary Learning [48.41770886055744]
  Federated Learning has seen an increased deployment in real-world scenarios recently.<n>The introduction of the so-called gradient inversion attacks has challenged its privacy-preserving properties.<n>We introduce SPEAR, which is based on a theoretical analysis of the gradients of linear layers with ReLU activations.<n>Our new attack, SPEAR++, retains all desirable properties of SPEAR, such as robustness to DP noise and FedAvg aggregation.
  arXiv  Detail & Related papers  (2025-10-28T09:06:19Z)
• VIKING: Deep variational inference with stochastic projections [48.946143517489496]
  Variational mean field approximations tend to struggle with contemporary overparametrized deep neural networks.<n>We propose a simple variational family that considers two independent linear subspaces of the parameter space.<n>This allows us to build a fully-correlated approximate posterior reflecting the overparametrization.
  arXiv  Detail & Related papers  (2025-10-27T15:38:35Z)
• Preserving instance continuity and length in segmentation through connectivity-aware loss computation [0.29656637520758655]
  We propose two novel loss functions, Negative Centerline Loss and Simplified Topology Loss, that help preserve connectivity of output instances.<n>We evaluate our approach on a 3D light-sheet fluorescence microscopy dataset of axon initial segments (AIS), a task prone to discontinuity due to signal dropout.
  arXiv  Detail & Related papers  (2025-09-03T09:08:47Z)
• Description of the Training Process of Neural Networks via Ergodic Theorem : Ghost nodes [3.637162892228131]
  We present a unified framework for understanding and accelerating deep neural networks via training gradient descent (SGD)<n>We introduce a practical diagnostic, the running estimate of the largest Lyapunov exponent, which distinguishes genuine convergence toward stablers.<n>We propose a ghost category extension for standard classifiers that adds auxiliary ghost output nodes so the model gains extra descent directions.
  arXiv  Detail & Related papers  (2025-07-01T17:54:35Z)
• On the Dynamics Under the Unhinged Loss and Beyond [104.49565602940699]
  We introduce the unhinged loss, a concise loss function, that offers more mathematical opportunities to analyze closed-form dynamics. The unhinged loss allows for considering more practical techniques, such as time-vary learning rates and feature normalization.
  arXiv  Detail & Related papers  (2023-12-13T02:11:07Z)
• Latent Point Collapse on a Low Dimensional Embedding in Deep Neural Network Classifiers [0.0]
  We propose a method to induce the collapse of latent representations belonging to the same class into a single point.<n>The proposed approach is straightforward to implement and yields substantial improvements in discnative feature embeddings.
  arXiv  Detail & Related papers  (2023-10-12T11:16:57Z)
• Collapsed Inference for Bayesian Deep Learning [36.1725075097107]
  We introduce a novel collapsed inference scheme that performs Bayesian model averaging using collapsed samples. A collapsed sample represents uncountably many models drawn from the approximate posterior. Our proposed use of collapsed samples achieves a balance between scalability and accuracy.
  arXiv  Detail & Related papers  (2023-06-16T08:34:42Z)
• Learning an Invertible Output Mapping Can Mitigate Simplicity Bias in Neural Networks [66.76034024335833]
  We investigate why diverse/ complex features are learned by the backbone, and their brittleness is due to the linear classification head relying primarily on the simplest features. We propose Feature Reconstruction Regularizer (FRR) to ensure that the learned features can be reconstructed back from the logits. We demonstrate up to 15% gains in OOD accuracy on the recently introduced semi-synthetic datasets with extreme distribution shifts.
  arXiv  Detail & Related papers  (2022-10-04T04:01:15Z)
• An Unconstrained Layer-Peeled Perspective on Neural Collapse [20.75423143311858]
  We introduce a surrogate model called the unconstrained layer-peeled model (ULPM) We prove that gradient flow on this model converges to critical points of a minimum-norm separation problem exhibiting neural collapse in its global minimizer. We show that our results also hold during the training of neural networks in real-world tasks when explicit regularization or weight decay is not used.
  arXiv  Detail & Related papers  (2021-10-06T14:18:47Z)
• DeepSplit: Scalable Verification of Deep Neural Networks via Operator Splitting [70.62923754433461]
  Analyzing the worst-case performance of deep neural networks against input perturbations amounts to solving a large-scale non- optimization problem. We propose a novel method that can directly solve a convex relaxation of the problem to high accuracy, by splitting it into smaller subproblems that often have analytical solutions.
  arXiv  Detail & Related papers  (2021-06-16T20:43:49Z)
• Online Limited Memory Neural-Linear Bandits with Likelihood Matching [53.18698496031658]
  We study neural-linear bandits for solving problems where both exploration and representation learning play an important role. We propose a likelihood matching algorithm that is resilient to catastrophic forgetting and is completely online.
  arXiv  Detail & Related papers  (2021-02-07T14:19:07Z)

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.