Abstract | ||
---|---|---|
We propose an approach for improving sequence modeling based on autoregressive normalizing flows. Each autoregressive transform, acting across time, serves as a moving frame of reference, removing temporal correlations and simplifying the modeling of higher-level dynamics. This technique provides a simple, general-purpose method for improving sequence modeling, with connections to existing and classical techniques. We demonstrate the proposed approach both with standalone flow-based models and as a component within sequential latent variable models. Results are presented on three benchmark video datasets and three other time series datasets, where autoregressive flow-based dynamics improve log-likelihood performance over baseline models. Finally, we illustrate the decorrelation and improved generalization properties of using flow-based dynamics. |
Year | DOI | Venue |
---|---|---|
2022 | 10.1007/s10994-021-06092-6 | Machine Learning |
Keywords | DocType | Volume |
Autoregressive flows, Latent variable models, Sequence modeling | Conference | 111 |
Issue | ISSN | Citations |
4 | 0885-6125 | 0 |
PageRank | References | Authors |
0.34 | 6 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Joseph Marino | 1 | 70 | 11.35 |
Lei Chen | 2 | 34 | 4.05 |
Jiawei He | 3 | 8 | 2.86 |
Mandt, Stephan | 4 | 128 | 19.55 |