<ul><li>Research on building interpretable neural architectures like CoFrNets has been relatively sparse.</li><li>CoFrNets is a novel neural architecture inspired by continued fractions, known for their attractive properties in number theory.</li><li>CoFrNets can be efficiently trained and interpreted due to their specific functional form, serving as universal approximators.</li><li>Experiments on synthetic and real datasets show that CoFrNets are competitive or superior to other interpretable models and multilayer perceptrons.</li></ul>

CoFrNets: Interpretable Neural Architecture Inspired by Continued Fractions

Discover more