Denoising diffusion probabilistic models (DDPMs) have shown impressive results on sequence generation by iteratively corrupting each example and then learning to map corrupted versions back to the original. However, previous work has largely focused on in-place corruption, adding noise to each pixel or token individually while keeping their locations the same. In this work, we consider a broader class of corruption processes and denoising models over sequence data that can insert and delete elements, while still being efficient to train and sample from. We demonstrate that these models outperform standard in-place models on an arithmetic sequence task, and that when trained on the text8 dataset they can be used to fix spelling errors without any fine-tuning.
@article{arxiv.2107.07675,
title = {Beyond In-Place Corruption: Insertion and Deletion In Denoising Probabilistic Models},
author = {Daniel D. Johnson and Jacob Austin and Rianne van den Berg and Daniel Tarlow},
journal= {arXiv preprint arXiv:2107.07675},
year = {2021}
}
Comments
Accepted at the ICML 2021 Workshop on Invertible Neural Networks, Normalizing Flows, and Explicit Likelihood Models (poster)