Learning Conditional Invariance through Cycle Consistency

Abstract

Identifying meaningful and independent factors of variation in a dataset is a challenging learning task frequently addressed by means of deep latent variable models. This task can be viewed as learning symmetry transformations preserving the value of a chosen property along latent dimensions. However, existing approaches exhibit severe drawbacks in enforcing the invariance property in the latent space. We address these shortcomings with a novel approach to cycle consistency. Our method involves two separate latent subspaces for the target property and the remaining input information, respectively. In order to enforce invariance as well as sparsity in the latent space, we incorporate semantic knowledge by using cycle consistency constraints relying on property side information. The proposed method is based on the deep information bottleneck and, in contrast to other approaches, allows using continuous target properties and provides inherent model selection capabilities. We demonstrate on synthetic and molecular data that our approach identifies more meaningful factors which lead to sparser and more interpretable models with improved invariance properties.

16 pages, 3 figures, published at the DAGM German Conference on Pattern Recognition, Sep. 28 - Oct. 1, 2021

References in corpus (8)

• Semi-Supervised Learning with Deep Generative Models
• Multi-Level Variational Autoencoder: Learning Disentangled Representations from Grouped Observations
• Learning Latent Subspaces in Variational Autoencoders
• 3DMolNet: A Generative Network for Molecular Structures
• On the Difference Between the Information Bottleneck and the Deep Information Bottleneck
• Causal Compression
• DualDis: Dual-Branch Disentangling with Adversarial Learning
• Causal learning with sufficient statistics: an information bottleneck approach