Learning the Irreducible Representations of Commutative Lie Groups

Abstract

We present a new probabilistic model of compact commutative Lie groups that produces invariant-equivariant and disentangled representations of data. To define the notion of disentangling, we borrow a fundamental principle from physics that is used to derive the elementary particles of a system from its symmetries. Our model employs a newfound Bayesian conjugacy relation that enables fully tractable probabilistic inference over compact commutative Lie groups -- a class that includes the groups that describe the rotation and cyclic translation of images. We train the model on pairs of transformed image patches, and show that the learned invariant representation is highly effective for classification.

Cited by in corpus (26)

• Disentangling by Factorising
• Group Equivariant Convolutional Networks
• 3D Steerable CNNs: Learning Rotationally Equivariant Features in Volumetric Data
• Exploiting Cyclic Symmetry in Convolutional Neural Networks
• Are Disentangled Representations Helpful for Abstract Visual Reasoning?
• Disentangling Factors of Variation Using Few Labels
• Intertwiners between Induced Representations (with Applications to the Theory of Equivariant Neural Networks)
• Warped Convolutions: Efficient Invariance to Spatial Transformations
• Automatic Symmetry Discovery with Lie Algebra Convolutional Network
• Theory and Evaluation Metrics for Learning Disentangled Representations
• Correspondence-Free Point Cloud Registration with SO(3)-Equivariant Implicit Shape Representations
• Harmonic Exponential Families on Manifolds
• Learning Group Structure and Disentangled Representations of Dynamical Environments
• On the Fairness of Disentangled Representations
• Disentangled Representations in Neural Models
• Topographic VAEs learn Equivariant Capsules
• Disentangling by Subspace Diffusion
• Measuring the Biases and Effectiveness of Content-Style Disentanglement
• WeLa-VAE: Learning Alternative Disentangled Representations Using Weak Labels
• Disentangling images with Lie group transformations and sparse coding
• Disentanglement Learning via Topology
• A Metric for Linear Symmetry-Based Disentanglement
• FILTRA: Rethinking Steerable CNN by Filter Transform
• Quantifying and Learning Linear Symmetry-Based Disentanglement
• Learning Identity-Preserving Transformations on Data Manifolds
• Joint Estimation of Image Representations and their Lie Invariants