Saved in:
Bibliographic Details
Main Authors: Powell, Brian P., Caraballo-Vega, Jordan A., Carroll, Mark L., Maxwell, Thomas, Ptak, Andrew, Olmschenk, Greg, Martinez-Palomera, Jorge
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2512.10817
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915669630517248
author Powell, Brian P.
Caraballo-Vega, Jordan A.
Carroll, Mark L.
Maxwell, Thomas
Ptak, Andrew
Olmschenk, Greg
Martinez-Palomera, Jorge
author_facet Powell, Brian P.
Caraballo-Vega, Jordan A.
Carroll, Mark L.
Maxwell, Thomas
Ptak, Andrew
Olmschenk, Greg
Martinez-Palomera, Jorge
contents We report the discovery that binary encoding allows neural networks to extrapolate periodic functions beyond their training bounds. We introduce Normalized Base-2 Encoding (NB2E) as a method for encoding continuous numerical values and demonstrate that, using this input encoding, vanilla multi-layer perceptrons (MLP) successfully extrapolate diverse periodic signals without prior knowledge of their functional form. Internal activation analysis reveals that NB2E induces bit-phase representations, enabling MLPs to learn and extrapolate signal structure independently of position.
format Preprint
id arxiv_https___arxiv_org_abs_2512_10817
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Extrapolation of Periodic Functions Using Binary Encoding of Continuous Numerical Values
Powell, Brian P.
Caraballo-Vega, Jordan A.
Carroll, Mark L.
Maxwell, Thomas
Ptak, Andrew
Olmschenk, Greg
Martinez-Palomera, Jorge
Machine Learning
Artificial Intelligence
Computer Vision and Pattern Recognition
We report the discovery that binary encoding allows neural networks to extrapolate periodic functions beyond their training bounds. We introduce Normalized Base-2 Encoding (NB2E) as a method for encoding continuous numerical values and demonstrate that, using this input encoding, vanilla multi-layer perceptrons (MLP) successfully extrapolate diverse periodic signals without prior knowledge of their functional form. Internal activation analysis reveals that NB2E induces bit-phase representations, enabling MLPs to learn and extrapolate signal structure independently of position.
title Extrapolation of Periodic Functions Using Binary Encoding of Continuous Numerical Values
topic Machine Learning
Artificial Intelligence
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2512.10817