Difference-in-Differences: Bridging Normalization and Disentanglement in PG-GAN

TL;DR

This paper introduces a difference-in-difference framework to analyze how pixel normalization in PG-GAN causes entanglement, revealing that in-painting during ablation leads to object entanglement, with the relation between units and objects being crucial.

Contribution

It proposes a novel DID-based experimental framework to uncover the mechanisms behind GAN entanglement, specifically linking pixel normalization to object entanglement during transformations.

Findings

Pixel normalization causes object entanglement via in-painting.

The unit-object relation influences the entanglement process.

The DID framework provides a solid, explainable analysis of the mechanism.

Abstract

What mechanisms causes GAN's entanglement? Although developing disentangled GAN has attracted sufficient attention, it is unclear how entanglement is originated by GAN transformation. We in this research propose a difference-in-difference (DID) counterfactual framework to design experiments for analyzing the entanglement mechanism in on of the Progressive-growing GAN (PG-GAN). Our experiment clarify the mechanisms how pixel normalization causes PG-GAN entanglement during a input-unit-ablation transformation. We discover that pixel normalization causes object entanglement by in-painting the area occupied by ablated objects. We also discover the unit-object relation determines whether and how pixel normalization causes objects entanglement. Our DID framework theoretically guarantees that the mechanisms that we discover is solid, explainable and comprehensively.