Alternators With Noise Models

TL;DR

Alternator++ is a novel time-series modeling framework that explicitly incorporates noise models to improve flexibility and performance in tasks like density estimation, imputation, and forecasting.

Contribution

It introduces Alternator++, a new model that enhances traditional Alternators by explicitly modeling and matching noise trajectories, improving performance on time-series tasks.

Findings

Outperforms baselines like Mambas, ScoreGrad, and Dyffusion.

Effective in density estimation, imputation, and forecasting.

Enhances model flexibility through explicit noise modeling.

Abstract

Alternators have recently been introduced as a framework for modeling time-dependent data. They often outperform other popular frameworks, such as state-space models and diffusion models, on challenging time-series tasks. This paper introduces a new Alternator model, called Alternator++, which enhances the flexibility of traditional Alternators by explicitly modeling the noise terms used to sample the latent and observed trajectories, drawing on the idea of noise models from the diffusion modeling literature. Alternator++ optimizes the sum of the Alternator loss and a noise-matching loss. The latter forces the noise trajectories generated by the two noise models to approximate the noise trajectories that produce the observed and latent trajectories. We demonstrate the effectiveness of Alternator++ in tasks such as density estimation, time series imputation, and forecasting, showing that it outperforms several strong baselines, including Mambas, ScoreGrad, and Dyffusion.