DeepAI
Log In Sign Up

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

10/16/2020
by   Xiao Liu, et al.
0

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.

READ FULL TEXT

page 2

page 5

page 7

page 8

02/13/2018

Homological analysis of multi-qubit entanglement

We propose the usage of persistent homologies to characterize multiparti...
11/18/2019

Quantifying the unextendibility of entanglement

The unextendibility or monogamy of entangled states is a key property of...
11/25/2022

A Configurable Protocol for Quantum Entanglement Distribution to End Nodes

The primary task of a quantum repeater network is to deliver entanglemen...
05/04/2017

Pixel Normalization from Numeric Data as Input to Neural Networks

Text to image transformation for input to neural networks requires inter...
07/16/2019

Persistent homology analysis of multiqubit entanglement

We introduce a homology-based technique for the analysis of multiqubit s...
06/02/2021

A Topological Solution of Entanglement for Complex-shaped Parts in Robotic Bin-picking

This paper addresses the problem of picking up only one object at a time...
05/12/2014

A Study of Entanglement in a Categorical Framework of Natural Language

In both quantum mechanics and corpus linguistics based on vector spaces,...