Boxhead: A Dataset for Learning Hierarchical Representations

Abstract

Disentanglement is hypothesized to be beneficial towards a number of downstream tasks. However, a common assumption in learning disentangled representations is that the data generative factors are statistically independent. As current methods are almost solely evaluated on toy datasets where this ideal assumption holds, we investigate their performance in hierarchical settings, a relevant feature of real-world data. In this work, we introduce Boxhead, a dataset with hierarchically structured ground-truth generative factors. We use this novel dataset to evaluate the performance of state-of-the-art autoencoder-based disentanglement models and observe that hierarchical models generally outperform single-layer VAEs in terms of disentanglement of hierarchically arranged factors.

arXiv commentNeurIPS 2021 Workshop on Shared Visual Representations in Human and Machine Intelligence (SVRHM 2021)

References in corpus (24)

• InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets
• Open3D: A Modern Library for 3D Data Processing
• NVAE: A Deep Hierarchical Variational Autoencoder
• Ladder Variational Autoencoders
• Multi-Object Representation Learning with Iterative Variational Inference
• Flexibly Fair Representation Learning by Disentanglement
• Causality for Machine Learning
• Are Disentangled Representations Helpful for Abstract Visual Reasoning?
• BIVA: A Very Deep Hierarchy of Latent Variables for Generative Modeling
• Weakly-Supervised Disentanglement Without Compromises
• Life-Long Disentangled Representation Learning with Cross-Domain Latent Homologies
• Recurrent Independent Mechanisms
• Weakly Supervised Disentanglement with Guarantees
• Spatial Broadcast Decoder: A Simple Architecture for Learning Disentangled Representations in VAEs
• Variational Inference of Disentangled Latent Concepts from Unlabeled Observations
• Causal Consistency of Structural Equation Models
• Very Deep VAEs Generalize Autoregressive Models and Can Outperform Them on Images
• Unsupervised Discovery of Parts, Structure, and Dynamics
• Disentangled State Space Representations
• SplineNets: Continuous Neural Decision Graphs
• Progressive Learning and Disentanglement of Hierarchical Representations
• S2RMs: Spatially Structured Recurrent Modules
• Benchmarks, Algorithms, and Metrics for Hierarchical Disentanglement
• Generalization and Robustness Implications in Object-Centric Learning