On Linear Interpolation in the Latent Space of Deep Generative Models

05/08/2021
by   Mike Yan Michelis, et al.
15

The underlying geometrical structure of the latent space in deep generative models is in most cases not Euclidean, which may lead to biases when comparing interpolation capabilities of two models. Smoothness and plausibility of linear interpolations in latent space are associated with the quality of the underlying generative model. In this paper, we show that not all such interpolations are comparable as they can deviate arbitrarily from the shortest interpolation curve given by the geodesic. This deviation is revealed by computing curve lengths with the pull-back metric of the generative model, finding shorter curves than the straight line between endpoints, and measuring a non-zero relative length improvement on this straight line. This leads to a strategy to compare linear interpolations across two generative models. We also show the effect and importance of choosing an appropriate output space for computing shorter curves. For this computation we derive an extension of the pull-back metric.

READ FULL TEXT

page 3

page 4

page 6

page 7

page 8

page 9

research
12/26/2018

Latent Variable Modeling for Generative Concept Representations and Deep Generative Models

Latent representations are the essence of deep generative models and det...
research
10/06/2022

Content-Based Search for Deep Generative Models

The growing proliferation of pretrained generative models has made it in...
research
10/19/2020

Statistical guarantees for generative models without domination

In this paper, we introduce a convenient framework for studying (adversa...
research
08/07/2023

Generative Models for the Deformation of Industrial Shapes with Linear Geometric Constraints: model order and parameter space reductions

Real-world applications of computational fluid dynamics often involve th...
research
11/23/2021

A Contextual Latent Space Model: Subsequence Modulation in Melodic Sequence

Some generative models for sequences such as music and text allow us to ...
research
04/06/2019

Interpolation in generative models

We show how to construct smooth and realistic interpolations for generat...
research
01/20/2019

Data Interpolations in Deep Generative Models under Non-Simply-Connected Manifold Topology

Exploiting the deep generative model's remarkable ability of learning th...

Please sign up or login with your details

Forgot password? Click here to reset