Dimensionality compression and expansion in Deep Neural Networks

Abstract

Datasets such as images, text, or movies are embedded in high-dimensional spaces. However, in important cases such as images of objects, the statistical structure in the data constrains samples to a manifold of dramatically lower dimensionality. Learning to identify and extract task-relevant variables from this embedded manifold is crucial when dealing with high-dimensional problems. We find that neural networks are often very effective at solving this task and investigate why. To this end, we apply state-of-the-art techniques for intrinsic dimensionality estimation to show that neural networks learn low-dimensional manifolds in two phases: first, dimensionality expansion driven by feature generation in initial layers, and second, dimensionality compression driven by the selection of task-relevant features in later layers. We model noise generated by Stochastic Gradient Descent and show how this noise balances the dimensionality of neural representations by inducing an effective regularization term in the loss. We highlight the important relationship between low-dimensional compressed representations and generalization properties of the network. Our work contributes by shedding light on the success of deep neural networks in disentangling data in high-dimensional space while achieving good generalization. Furthermore, it invites new learning strategies focused on optimizing measurable geometric properties of learned representations, beginning with their intrinsic dimensionality.

arXiv commentSubmitted to NeurIPS 2019. First two authors contributed equally

References in corpus (12)

• Estimating the intrinsic dimension of datasets by a minimal neighborhood information
• Sensitivity and Generalization in Neural Networks: an Empirical Study
• Dropout Training as Adaptive Regularization
• Classification and Geometry of General Perceptual Manifolds
• Towards Understanding Generalization of Deep Learning: Perspective of Loss Landscapes
• Fisher-Rao Metric, Geometry, and Complexity of Neural Networks
• Energy-entropy competition and the effectiveness of stochastic gradient descent in machine learning
• Linear Readout of Object Manifolds
• Manifold Learning with Geodesic Minimal Spanning Trees
• DANCo: Dimensionality from Angle and Norm Concentration
• The role of dimensionality reduction in linear classification
• The Local Dimension of Deep Manifold

Cited by in corpus (11)

• Transient chaotic dimensionality expansion by recurrent networks
• Deep Nonparametric Regression on Approximate Manifolds: Non-Asymptotic Error Bounds with Polynomial Prefactors
• Emergence of Separable Manifolds in Deep Language Representations
• Weakly-correlated synapses promote dimension reduction in deep neural networks
• CNNs Avoid Curse of Dimensionality by Learning on Patches
• Inversion dynamics of class manifolds in deep learning reveals tradeoffs underlying generalisation
• Relative stability toward diffeomorphisms indicates performance in deep nets
• Perspective: A Phase Diagram for Deep Learning unifying Jamming, Feature Learning and Lazy Training
• Channel-Wise Early Stopping without a Validation Set via NNK Polytope Interpolation
• Internal representation dynamics and geometry in recurrent neural networks
• The Foes of Neural Network's Data Efficiency Among Unnecessary Input Dimensions