pith. sign in

arxiv: 2410.05050 · v2 · pith:MDY5JNFTnew · submitted 2024-10-07 · 💻 cs.LG · cs.AI· stat.ML

FreSh: Frequency Shifting for Accelerated Neural Representation Learning

classification 💻 cs.LG cs.AIstat.ML
keywords frequencymodelhyperparametersneuralembeddingsfreshhigh-frequencyinitial
0
0 comments X
read the original abstract

Implicit Neural Representations (INRs) have recently gained attention as a powerful approach for continuously representing signals such as images, videos, and 3D shapes using multilayer perceptrons (MLPs). However, MLPs are known to exhibit a low-frequency bias, limiting their ability to capture high-frequency details accurately. This limitation is typically addressed by incorporating high-frequency input embeddings or specialized activation layers. In this work, we demonstrate that these embeddings and activations are often configured with hyperparameters that perform well on average but are suboptimal for specific input signals under consideration, necessitating a costly grid search to identify optimal settings. Our key observation is that the initial frequency spectrum of an untrained model's output correlates strongly with the model's eventual performance on a given target signal. Leveraging this insight, we propose frequency shifting (or FreSh), a method that selects embedding hyperparameters to align the frequency spectrum of the model's initial output with that of the target signal. We show that this simple initialization technique improves performance across various neural representation methods and tasks, achieving results comparable to extensive hyperparameter sweeps but with only marginal computational overhead compared to training a single model with default hyperparameters.

This paper has not been read by Pith yet.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. JA-SIREN: Deterministic Initialization for Sinusoidal Networks via Spectral Matching

    cs.CV 2026-06 unverdicted novelty 7.0

    JA-SIREN derives deterministic weights for two-layer sinusoidal MLPs via DST and Jacobi-Anger expansion to match target spectra, yielding 67.18 dB PSNR on Kodak with zero variance.

  2. Implicit Neural Representations: A Signal Processing Perspective

    cs.CV 2026-04 unverdicted novelty 3.0

    INRs parameterize signals as neural networks to enable continuous representations, analytical differentiation, and adaptive approximation spaces that address spectral bias through specialized activations and structure...