Poster
in
Workshop: Geometry-grounded Representation Learning and Generative Modeling
All Roads Lead to Rome? Exploring Representational Similarities Between Latent Spaces of Generative Image Models
Charumathi Badrinath · Usha Bhalla · Alex Oesterling · Suraj Srinivas · Himabindu Lakkaraju
Keywords: [ Generative Models ] [ Latent Space ] [ Representations ]
Do different generative image models secretly learn similar underlying representations? We investigate this by measuring the latent space similarity of four different models: VAEs, GANs, Normalizing Flows (NFs), and Diffusion Models (DMs). Our methodology involves training linear maps between frozen latent spaces to "stitch" arbitrary pairs of encoders and decoders and measuring output-based and probe-based metrics on the resulting "stitched'' models. Our main findings are that linear maps between latent spaces of performant models preserve most visual information even when latent sizes differ; for CelebA models, gender is the most similarly represented probe-able attribute. Finally we show on an NF that latent space representations converge early in training.