Continued Fraction Neural Networks (CFNN) introduce a rational inductive bias that handles singularities with 10-100x fewer parameters than standard MLPs.
March 24, 2026
Original Paper
CFNN: Continued Fraction Neural Network
arXiv · 2603.20634
The Takeaway
By integrating continued fractions into the architecture, this paper provides a 'grey-box' model for scientific computing that is far more parameter-frugal and robust to noise than traditional neural networks when modeling complex physical functions.
From the abstract
Accurately characterizing non-linear functional manifolds with singularities is a fundamental challenge in scientific computing. While Multi-Layer Perceptrons (MLPs) dominate, their spectral bias hinders resolving high-curvature features without excessive parameters. We introduce Continued Fraction Neural Networks (CFNNs), integrating continued fractions with gradient-based optimization to provide a ``rational inductive bias.'' This enables capturing complex asymptotics and discontinuities with