Unveiling the Training Dynamics of ReLU Networks through a Linear Lens

• URL: http://arxiv.org/abs/2511.05628v1
• Date: Fri, 07 Nov 2025 02:59:51 GMT
• Title: Unveiling the Training Dynamics of ReLU Networks through a Linear Lens
• Authors: Longqing Ye,
• Abstract summary: We recast a multi-layer ReLU network into a single-layer linear model with input-dependent "effective weights"<n>For any given input sample, the activation pattern of ReLU units creates a unique computational path.<n>As training progresses, the effective weights corresponding to samples from the same class converge, while those from different classes diverge.
• Score: 0.0
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Deep neural networks, particularly those employing Rectified Linear Units (ReLU), are often perceived as complex, high-dimensional, non-linear systems. This complexity poses a significant challenge to understanding their internal learning mechanisms. In this work, we propose a novel analytical framework that recasts a multi-layer ReLU network into an equivalent single-layer linear model with input-dependent "effective weights". For any given input sample, the activation pattern of ReLU units creates a unique computational path, effectively zeroing out a subset of weights in the network. By composing the active weights across all layers, we can derive an effective weight matrix, $W_{\text{eff}}(x)$, that maps the input directly to the output for that specific sample. We posit that the evolution of these effective weights reveals fundamental principles of representation learning. Our work demonstrates that as training progresses, the effective weights corresponding to samples from the same class converge, while those from different classes diverge. By tracking the trajectories of these sample-wise effective weights, we provide a new lens through which to interpret the formation of class-specific decision boundaries and the emergence of semantic representations within the network.

Related papers

• IF2Net: Innately Forgetting-Free Networks for Continual Learning [49.57495829364827]
  Continual learning can incrementally absorb new concepts without interfering with previously learned knowledge. Motivated by the characteristics of neural networks, we investigated how to design an Innately Forgetting-Free Network (IF2Net) IF2Net allows a single network to inherently learn unlimited mapping rules without telling task identities at test time.
  arXiv  Detail & Related papers  (2023-06-18T05:26:49Z)
• ReLU Neural Networks with Linear Layers are Biased Towards Single- and Multi-Index Models [9.96121040675476]
  This manuscript explores how properties of functions learned by neural networks of depth greater than two layers affect predictions.<n>Our framework considers a family of networks of varying depths that all have the same capacity but different representation costs.
  arXiv  Detail & Related papers  (2023-05-24T22:10:12Z)
• Equivariant Architectures for Learning in Deep Weight Spaces [54.61765488960555]
  We present a novel network architecture for learning in deep weight spaces. It takes as input a concatenation of weights and biases of a pre-trainedvariant. We show how these layers can be implemented using three basic operations.
  arXiv  Detail & Related papers  (2023-01-30T10:50:33Z)
• Slimmable Networks for Contrastive Self-supervised Learning [69.9454691873866]
  Self-supervised learning makes significant progress in pre-training large models, but struggles with small models. We introduce another one-stage solution to obtain pre-trained small models without the need for extra teachers. A slimmable network consists of a full network and several weight-sharing sub-networks, which can be pre-trained once to obtain various networks.
  arXiv  Detail & Related papers  (2022-09-30T15:15:05Z)
• BiTAT: Neural Network Binarization with Task-dependent Aggregated Transformation [116.26521375592759]
  Quantization aims to transform high-precision weights and activations of a given neural network into low-precision weights/activations for reduced memory usage and computation. Extreme quantization (1-bit weight/1-bit activations) of compactly-designed backbone architectures results in severe performance degeneration. This paper proposes a novel Quantization-Aware Training (QAT) method that can effectively alleviate performance degeneration.
  arXiv  Detail & Related papers  (2022-07-04T13:25:49Z)
• Training invariances and the low-rank phenomenon: beyond linear networks [44.02161831977037]
  We show that when one trains a deep linear network with logistic or exponential loss on linearly separable data, the weights converge to rank-$1$ matrices. This is the first time a low-rank phenomenon is proven rigorously for nonlinear ReLU-activated feedforward networks. Our proof relies on a specific decomposition of the network into a multilinear function and another ReLU network whose weights are constant under a certain parameter directional convergence.
  arXiv  Detail & Related papers  (2022-01-28T07:31:19Z)
• Gone Fishing: Neural Active Learning with Fisher Embeddings [55.08537975896764]
  There is an increasing need for active learning algorithms that are compatible with deep neural networks. This article introduces BAIT, a practical representation of tractable, and high-performing active learning algorithm for neural networks.
  arXiv  Detail & Related papers  (2021-06-17T17:26:31Z)
• MLDS: A Dataset for Weight-Space Analysis of Neural Networks [0.0]
  We present MLDS, a new dataset consisting of thousands of trained neural networks with carefully controlled parameters. This dataset enables new insights into both model-to-model and model-to-training-data relationships.
  arXiv  Detail & Related papers  (2021-04-21T14:24:26Z)
• Stable Recovery of Entangled Weights: Towards Robust Identification of Deep Neural Networks from Minimal Samples [0.0]
  We introduce the so-called entangled weights, which compose weights of successive layers intertwined with suitable diagonal and invertible matrices depending on the activation functions and their shifts. We prove that entangled weights are completely and stably approximated by an efficient and robust algorithm. In terms of practical impact, our study shows that we can relate input-output information uniquely and stably to network parameters, providing a form of explainability.
  arXiv  Detail & Related papers  (2021-01-18T16:31:19Z)
• Statistical Mechanics of Deep Linear Neural Networks: The Back-Propagating Renormalization Group [4.56877715768796]
  We study the statistical mechanics of learning in Deep Linear Neural Networks (DLNNs) in which the input-output function of an individual unit is linear. We solve exactly the network properties following supervised learning using an equilibrium Gibbs distribution in the weight space. Our numerical simulations reveal that despite the nonlinearity, the predictions of our theory are largely shared by ReLU networks with modest depth.
  arXiv  Detail & Related papers  (2020-12-07T20:08:31Z)
• Neural networks with late-phase weights [66.72777753269658]
  We show that the solutions found by SGD can be further improved by ensembling a subset of the weights in late stages of learning. At the end of learning, we obtain back a single model by taking a spatial average in weight space.
  arXiv  Detail & Related papers  (2020-07-25T13:23:37Z)

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.