Latent Iterative Refinement for Modular Source Separation

TL;DR

This paper introduces a novel iterative refinement approach for source separation that enhances resource efficiency during training and inference by reusing processing blocks and dynamically adjusting iterations.

Contribution

It proposes reformulating source separation models as iterative latent mappings, enabling efficient training and inference through block reuse and dynamic iteration control.

Findings

Reduces memory requirements during training.

Improves resource efficiency with iterative processing.

Enables dynamic inference adjustments.

Abstract

Traditional source separation approaches train deep neural network models end-to-end with all the data available at once by minimizing the empirical risk on the whole training set. On the inference side, after training the model, the user fetches a static computation graph and runs the full model on some specified observed mixture signal to get the estimated source signals. Additionally, many of those models consist of several basic processing blocks which are applied sequentially. We argue that we can significantly increase resource efficiency during both training and inference stages by reformulating a model's training and inference procedures as iterative mappings of latent signal representations. First, we can apply the same processing block more than once on its output to refine the input signal and consequently improve parameter efficiency. During training, we can follow a block-wise procedure which enables a reduction on memory requirements. Thus, one can train a very complicated network structure using significantly less computation compared to end-to-end training. During inference, we can dynamically adjust how many processing blocks and iterations of a specific block an input signal needs using a gating module.