INR-Bench: A Unified Benchmark for Implicit Neural Representations in Multi-Domain Regression and Reconstruction
This work addresses the need for a unified benchmark to understand INR effectiveness and limitations for researchers in signal processing and neural representation fields, though it is incremental as it builds on existing NTK theory and model types.
The authors tackled the underexplored factors influencing Implicit Neural Representations (INRs) by introducing INR-Bench, a comprehensive benchmark with 56 Coordinate-MLP and 22 Coordinate-KAN variants evaluated across 9 multimodal tasks, providing a robust platform to analyze model architectures, positional encoding, and activation functions.
Implicit Neural Representations (INRs) have gained success in various signal processing tasks due to their advantages of continuity and infinite resolution. However, the factors influencing their effectiveness and limitations remain underexplored. To better understand these factors, we leverage insights from Neural Tangent Kernel (NTK) theory to analyze how model architectures (classic MLP and emerging KAN), positional encoding, and nonlinear primitives affect the response to signals of varying frequencies. Building on this analysis, we introduce INR-Bench, the first comprehensive benchmark specifically designed for multimodal INR tasks. It includes 56 variants of Coordinate-MLP models (featuring 4 types of positional encoding and 14 activation functions) and 22 Coordinate-KAN models with distinct basis functions, evaluated across 9 implicit multimodal tasks. These tasks cover both forward and inverse problems, offering a robust platform to highlight the strengths and limitations of different neural models, thereby establishing a solid foundation for future research. The code and dataset are available at https://github.com/lif314/INR-Bench.