Convergence Analysis of Flow Matching in Latent Space with Transformers

arXiv:2404.02538v1 Announce Type: cross
Abstract: We present theoretical convergence guarantees for ODE-based generative models, specifically flow matching. We use a pre-trained autoencoder network to map high-dimensional original inputs to a low-dimensional latent space, where a transformer network is trained to predict the velocity field of the transformation from a standard normal distribution to the target latent distribution. Our error analysis demonstrates the effectiveness of this approach, showing that the distribution of samples generated via estimated ODE flow converges to the …

abstract analysis arxiv autoencoder convergence cs.lg flow generative generative models inputs low map network space standard stat.ml transformation transformer transformer network transformers type