Multi-resolution enhancement for full-spectrum neural representations

TL;DR - Ni et al. introduce WIEN-INR, an implicit neural representation for compressing scientific data in a multiscale wavelet domain. The approach aims to improve compression while retaining fine-grained details and signal fidelity.

• Represents scientific data using an implicit neural representation.
• Operates across multiple resolutions in the wavelet domain.
• Targets both compact storage and preservation of fine details.
• Focuses on maintaining signal fidelity during compression.