Efficient Data Mosaicing with Simulation-based Inference

TL;DR

This paper presents a fast, simulation-based inference algorithm for data mosaicing that efficiently approximates target data fragments using source data, applicable to audio and image processing.

Contribution

The paper introduces a novel, parallelizable algorithm for data mosaicing leveraging simulation-based inference, enabling practical and efficient approximation of target data.

Findings

Effective in audio mosaicing tasks

Effective in image mosaicing tasks

Fast approximate posterior inference achieved

Abstract

We introduce an efficient algorithm for general data mosaicing, based on the simulation-based inference paradigm. Our algorithm takes as input a target datum, source data, and partitions of the target and source data into fragments, learning distributions over averages of fragments of the source data such that samples from those distributions approximate fragments of the target datum. We utilize a model that can be trivially parallelized in conjunction with the latest advances in efficient simulation-based inference in order to find approximate posteriors fast enough for use in practical applications. We demonstrate our technique is effective in both audio and image mosaicing problems.