AI RESEARCH

Puppet-CNN: Continuous Parameter Dynamics for Input-Adaptive Convolutional Networks

arXiv CS.LG

ArXi:2411.12876v2 Announce Type: replace Modern convolutional neural networks (CNNs) organize computation as a discrete stack of layers whose parameters are independently d and learned, with the number of layers fixed as an architectural hyperparameter. In this work, we explore an alternative perspective: can network parameterization itself be modeled as a continuous dynamical system? We