Extrapolation of Periodic Functions Using Binary Encoding of Continuous Numerical Values
By: Brian P. Powell , Jordan A. Caraballo-Vega , Mark L. Carroll and more
Potential Business Impact:
Makes computers learn patterns they haven't seen.
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.
Similar Papers
Neural Functions for Learning Periodic Signal
Machine Learning (CS)
Makes computer models better at predicting future patterns.
BEP: A Binary Error Propagation Algorithm for Binary Neural Networks Training
Machine Learning (CS)
Trains AI using only simple math for speed.
From Taylor Series to Fourier Synthesis: The Periodic Linear Unit
Machine Learning (CS)
Makes AI smarter with fewer computer parts.