Latent Autoregressive Source Separation

Autoregressive models have achieved impressive results over a wide range of domains in terms of generation quality and downstream task performance. In the continuous domain, a key factor behind this success is the usage of quantized latent spaces (e.g., obtained via VQ-VAE autoencoders), which al low for dimensionality reduction and faster inference times. However, using existing pre-trained models to perform new non-trivial tasks is difficult since it requires additional fine tuning or extensive training to elicit prompting. This paper introduces LASS as a way to perform vector-quantized La tent Autoregressive Source Separation (i.e., de-mixing an in put signal into its constituent sources) without requiring ad ditional gradient-based optimization or modifications of ex isting models. Our separation method relies on the Bayesian formulation in which the autoregressive models are the pri ors, and a discrete (non-parametric) likelihood function is constructed by performing frequency counts over latent sums of addend tokens. We test our method on images and au dio with several sampling strategies (e.g., ancestral, beam search) showing competitive results with existing approaches in terms of separation quality while offering at the same time significant speedups in terms of inference time and scalability to higher dimensional data.