Learning reshapes power-law anisotropy in internal representations

• URL: http://arxiv.org/abs/2608.15239v1
• Date: Sat, 15 Aug 2026 13:56:28 GMT
• Title: Learning reshapes power-law anisotropy in internal representations
• Abstract summary: Power-law anisotropy in internal representations has been observed across a wide range of biological and artificial neural systems.<n>We show that, in the feature-learning regime, the local power-law exponent of the internal-representation evolves nonmonotonically over the course of training.<n>These results suggest a general mechanism by which the dynamic interaction between input statistics and task structure gives rise to power-law internal representations.
• Score: 0.03437656066916039
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Power-law anisotropy in internal representations has been observed across a wide range of biological and artificial neural systems, from state-of-the-art language models to the mouse cerebral cortex. This anisotropy is a key geometric property of high-dimensional information processing and underlies a variety of theoretical analyses. However, the mechanism by which it emerges from input structure and task-driven learning has remained unclear. Here, we characterize this formation process by exactly solving the learning dynamics of a wide two-layer linear neural network in a teacher--student setting with power-law input and teacher structures. We show that, in the feature-learning regime, the local power-law exponent of the internal-representation spectrum evolves nonmonotonically over the course of training and exhibits up to four distinct asymptotic regimes across modes and training times. By contrast, in the lazy regime, the exponent remains essentially unchanged. We further demonstrate numerically that similar exponent dynamics arise in more realistic nonlinear networks. Together, these results suggest a general mechanism by which the dynamic interaction between input statistics and task structure gives rise to power-law internal representations.

Related papers

• Invariant Learning Dynamics of Transformers in Inductive Reasoning Tasks [50.79340930913805]
  We study a class of inductive tasks that unifies several synthetic tasks known in the literature, including in-context n-grams and multi-hop reasoning.<n>By casting circuit formation as a low-dimensional dynamical phenomenon, we take a step toward a predictive theory of how Transformers learn.
  arXiv  Detail & Related papers  (2026-07-13T17:56:11Z)
• Interpreting learning dynamics of autoencoders: Transient scaling and emerging concepts of the Ising model [0.0]
  unsupervised autoencoders trained on microscopic spin configurations from the Ising model learn macroscopic, theory-relevant variables.<n>We quantify learning across multiple spatial scales and reveal two distinct dynamical regimes.<n>A dynamical viewpoint of learning is established in which intrinsic properties expose the effects of forced changes in representation.
  arXiv  Detail & Related papers  (2026-07-11T12:31:34Z)
• Kirchhoff-Inspired Neural Networks for Evolving High-Order Perception [51.28469959548178]
  We propose the Kirchhoff-Inspired Neural Network (KINN), a state-variable-based network architecture constructed based on Kirchhoff's current law.<n>KINN derives numerically stable state updates from fundamental ordinary differential equations, enabling the explicit decoupling and encoding of higher-order evolutionary components.
  arXiv  Detail & Related papers  (2026-03-25T06:17:31Z)
• Hierarchical Latent Structures in Data Generation Process Unify Mechanistic Phenomena across Scale [50.33952894976367]
  We use context-free grammars to generate synthetic corpora that are faithful and computationally efficient proxies for web-scale text corpora.<n>We investigate the emergence of three mechanistic phenomena: induction heads, function vectors, and the Hydra effect, under our designed data generation process.<n>Our findings suggest that hierarchical structures in the data generation process serve as the X-factor in explaining the emergence of these phenomena.
  arXiv  Detail & Related papers  (2026-02-04T14:49:13Z)
• KPFlow: An Operator Perspective on Dynamic Collapse Under Gradient Descent Training of Recurrent Networks [9.512147747894026]
  We show how a gradient flow can be decomposed into a product that involves two operators.<n>We show how their interplay gives rise to low-dimensional latent dynamics under GD.<n>For multi-task training, we show that the operators can be used to measure how objectives relevant to individual sub-tasks align.
  arXiv  Detail & Related papers  (2025-07-08T20:33:15Z)
• From Lazy to Rich: Exact Learning Dynamics in Deep Linear Networks [47.13391046553908]
  In artificial networks, the effectiveness of these models relies on their ability to build task specific representation.<n>Prior studies highlight that different initializations can place networks in either a lazy regime, where representations remain static, or a rich/feature learning regime, where representations evolve dynamically.<n>These solutions capture the evolution of representations and the Neural Kernel across the spectrum from the rich to the lazy regimes.
  arXiv  Detail & Related papers  (2024-09-22T23:19:04Z)
• Mechanistic Neural Networks for Scientific Machine Learning [58.99592521721158]
  We present Mechanistic Neural Networks, a neural network design for machine learning applications in the sciences. It incorporates a new Mechanistic Block in standard architectures to explicitly learn governing differential equations as representations. Central to our approach is a novel Relaxed Linear Programming solver (NeuRLP) inspired by a technique that reduces solving linear ODEs to solving linear programs.
  arXiv  Detail & Related papers  (2024-02-20T15:23:24Z)
• A Waddington landscape for prototype learning in generalized Hopfield networks [0.0]
  We study the learning dynamics of Generalized Hopfield networks. We observe a strong resemblance to the canalized, or low-dimensional, dynamics of cells as they differentiate.
  arXiv  Detail & Related papers  (2023-12-04T21:28:14Z)
• Decomposed Linear Dynamical Systems (dLDS) for learning the latent components of neural dynamics [6.829711787905569]
  We propose a new decomposed dynamical system model that represents complex non-stationary and nonlinear dynamics of time series data. Our model is trained through a dictionary learning procedure, where we leverage recent results in tracking sparse vectors over time. In both continuous-time and discrete-time instructional examples we demonstrate that our model can well approximate the original system.
  arXiv  Detail & Related papers  (2022-06-07T02:25:38Z)
• A duality connecting neural network and cosmological dynamics [0.0]
  We show that the dynamics of neural networks trained with gradient descent and the dynamics of scalar fields in a flat, vacuum energy dominated Universe are structurally related. This duality provides the framework for synergies between these systems, to understand and explain neural network dynamics.
  arXiv  Detail & Related papers  (2022-02-22T19:00:01Z)
• Data-driven emergence of convolutional structure in neural networks [83.4920717252233]
  We show how fully-connected neural networks solving a discrimination task can learn a convolutional structure directly from their inputs. By carefully designing data models, we show that the emergence of this pattern is triggered by the non-Gaussian, higher-order local structure of the inputs.
  arXiv  Detail & Related papers  (2022-02-01T17:11:13Z)
• Recurrent Neural Network Learning of Performance and Intrinsic Population Dynamics from Sparse Neural Data [77.92736596690297]
  We introduce a novel training strategy that allows learning not only the input-output behavior of an RNN but also its internal network dynamics. We test the proposed method by training an RNN to simultaneously reproduce internal dynamics and output signals of a physiologically-inspired neural model. Remarkably, we show that the reproduction of the internal dynamics is successful even when the training algorithm relies on the activities of a small subset of neurons.
  arXiv  Detail & Related papers  (2020-05-05T14:16:54Z)

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.