The Tunnel Effect: Building Data Representations in Deep Neural Networks

• URL: http://arxiv.org/abs/2305.19753v2
• Date: Mon, 30 Oct 2023 12:41:27 GMT
• Title: The Tunnel Effect: Building Data Representations in Deep Neural Networks
• Abstract summary: This paper shows that sufficiently deep networks split into two distinct parts that contribute to the resulting data representations differently. The initial layers create linearly-separable representations, while the subsequent layers compress these representations and have a minimal impact on the overall performance.
• Score: 32.46906832880099
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Deep neural networks are widely known for their remarkable effectiveness across various tasks, with the consensus that deeper networks implicitly learn more complex data representations. This paper shows that sufficiently deep networks trained for supervised image classification split into two distinct parts that contribute to the resulting data representations differently. The initial layers create linearly-separable representations, while the subsequent layers, which we refer to as \textit{the tunnel}, compress these representations and have a minimal impact on the overall performance. We explore the tunnel's behavior through comprehensive empirical studies, highlighting that it emerges early in the training process. Its depth depends on the relation between the network's capacity and task complexity. Furthermore, we show that the tunnel degrades out-of-distribution generalization and discuss its implications for continual learning.

Related papers

• Provable Learning of Random Hierarchy Models and Hierarchical Shallow-to-Deep Chaining [58.69016084278948]
  We consider a hierarchical context-free grammar introduced by arXiv:2307.02129 and conjectured to separate deep and shallow networks.<n>We prove that, under mild conditions, a deep convolutional network can be efficiently trained to learn this function class.
  arXiv  Detail & Related papers  (2026-01-27T16:19:54Z)
• Revisiting Deep Information Propagation: Fractal Frontier and Finite-size Effects [16.42026346710591]
  We study information propagation in randomly neural networks with finite width and reveal that the boundary between ordered and chaotic regimes exhibits a fractal structure.<n>This shows the fundamental complexity of neural network dynamics, in a setting that is independent of input data and optimization.
  arXiv  Detail & Related papers  (2025-08-05T08:49:24Z)
• Statistical Physics of Deep Neural Networks: Generalization Capability, Beyond the Infinite Width, and Feature Learning [0.0]
  This thesis applies physics-based insights to understand Deep Neural Networks (DNNs) By understanding when a network must learn data structure, it sheds light on fostering meaningful internal representations.
  arXiv  Detail & Related papers  (2025-01-31T16:43:57Z)
• Deep Learning Through A Telescoping Lens: A Simple Model Provides Empirical Insights On Grokking, Gradient Boosting & Beyond [61.18736646013446]
  In pursuit of a deeper understanding of its surprising behaviors, we investigate the utility of a simple yet accurate model of a trained neural network. Across three case studies, we illustrate how it can be applied to derive new empirical insights on a diverse range of prominent phenomena.
  arXiv  Detail & Related papers  (2024-10-31T22:54:34Z)
• Learning to Compress: Local Rank and Information Compression in Deep Neural Networks [2.483371671716944]
  Deep neural networks tend to exhibit a bias toward low-rank solutions during training, implicitly learning low-dimensional feature representations. We introduce the concept of local rank as a measure of feature manifold dimensionality and demonstrate that this rank decreases during the final phase of training. We argue that networks that reduce the rank of their learned representations also compress mutual information between inputs and intermediate layers.
  arXiv  Detail & Related papers  (2024-10-10T07:53:34Z)
• Understanding Deep Representation Learning via Layerwise Feature Compression and Discrimination [33.273226655730326]
  We show that each layer of a deep linear network progressively compresses within-class features at a geometric rate and discriminates between-class features at a linear rate. This is the first quantitative characterization of feature evolution in hierarchical representations of deep linear networks.
  arXiv  Detail & Related papers  (2023-11-06T09:00:38Z)
• Neural Collapse in the Intermediate Hidden Layers of Classification Neural Networks [0.0]
  (NC) gives a precise description of the representations of classes in the final hidden layer of classification neural networks. In the present paper, we provide the first comprehensive empirical analysis of the emergence of (NC) in the intermediate hidden layers.
  arXiv  Detail & Related papers  (2023-08-05T01:19:38Z)
• Complexity of Representations in Deep Learning [2.0219767626075438]
  We analyze the effectiveness of the learned representations in separating the classes from a data complexity perspective. We show how the data complexity evolves through the network, how it changes during training, and how it is impacted by the network design and the availability of training samples.
  arXiv  Detail & Related papers  (2022-09-01T15:20:21Z)
• Dynamic Inference with Neural Interpreters [72.90231306252007]
  We present Neural Interpreters, an architecture that factorizes inference in a self-attention network as a system of modules. inputs to the model are routed through a sequence of functions in a way that is end-to-end learned. We show that Neural Interpreters perform on par with the vision transformer using fewer parameters, while being transferrable to a new task in a sample efficient manner.
  arXiv  Detail & Related papers  (2021-10-12T23:22:45Z)
• What can linearized neural networks actually say about generalization? [67.83999394554621]
  In certain infinitely-wide neural networks, the neural tangent kernel (NTK) theory fully characterizes generalization. We show that the linear approximations can indeed rank the learning complexity of certain tasks for neural networks. Our work provides concrete examples of novel deep learning phenomena which can inspire future theoretical research.
  arXiv  Detail & Related papers  (2021-06-12T13:05:11Z)
• Redundant representations help generalization in wide neural networks [71.38860635025907]
  We study the last hidden layer representations of various state-of-the-art convolutional neural networks. We find that if the last hidden representation is wide enough, its neurons tend to split into groups that carry identical information, and differ from each other only by statistically independent noise.
  arXiv  Detail & Related papers  (2021-06-07T10:18:54Z)
• A neural anisotropic view of underspecification in deep learning [60.119023683371736]
  We show that the way neural networks handle the underspecification of problems is highly dependent on the data representation. Our results highlight that understanding the architectural inductive bias in deep learning is fundamental to address the fairness, robustness, and generalization of these systems.
  arXiv  Detail & Related papers  (2021-04-29T14:31:09Z)
• Variational Structured Attention Networks for Deep Visual Representation Learning [49.80498066480928]
  We propose a unified deep framework to jointly learn both spatial attention maps and channel attention in a principled manner. Specifically, we integrate the estimation and the interaction of the attentions within a probabilistic representation learning framework. We implement the inference rules within the neural network, thus allowing for end-to-end learning of the probabilistic and the CNN front-end parameters.
  arXiv  Detail & Related papers  (2021-03-05T07:37:24Z)
• Learning Connectivity of Neural Networks from a Topological Perspective [80.35103711638548]
  We propose a topological perspective to represent a network into a complete graph for analysis. By assigning learnable parameters to the edges which reflect the magnitude of connections, the learning process can be performed in a differentiable manner. This learning process is compatible with existing networks and owns adaptability to larger search spaces and different tasks.
  arXiv  Detail & Related papers  (2020-08-19T04:53:31Z)
• Towards Deeper Graph Neural Networks [63.46470695525957]
  Graph convolutions perform neighborhood aggregation and represent one of the most important graph operations. Several recent studies attribute this performance deterioration to the over-smoothing issue. We propose Deep Adaptive Graph Neural Network (DAGNN) to adaptively incorporate information from large receptive fields.
  arXiv  Detail & Related papers  (2020-07-18T01:11:14Z)
• Untangling in Invariant Speech Recognition [17.996356271398295]
  We study how information is untangled within neural networks trained to recognize speech. We observe speaker-specific nuisance variations are discarded by the network's hierarchy, whereas task-relevant properties are untangled in later layers. We find that the deep representations carry out significant temporal untangling by efficiently extracting task-relevant features at each time step of the computation.
  arXiv  Detail & Related papers  (2020-03-03T20:48:43Z)

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.